Qwen: Qwen3.5-35B-A3B

The Qwen3.5 Series 35B-A3B is a native vision-language model designed with a hybrid architecture that integrates linear attention mechanisms and a sparse mixture-of-experts model, achieving higher in…

qwen/qwen3.5-35b-a3b-20260224

The Qwen3.5 Series 35B-A3B is a native vision-language model designed with a hybrid architecture that integrates linear attention mechanisms and a sparse mixture-of-experts model, achieving higher inference efficiency. Its overall performance is comparable to that of the Qwen3.5-27B.

Specifications & capabilities

Input
text, image, video
Output
text
Context window · tokens
262144
Published model ID
Qwen3.5-35B-A3B
Capabilities
multimodal, promptCaching, reasoning, structuredOutputs, toolCalling, visionInput
Description source ↗

Evaluations

1–18 / 18

OpenRouter Benchmarks

76.333%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Darkbloom
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
1
sample size
500
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
10
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
tau_bench_verified_airline
name
tau2-bench Verified Airline
record_kind
aggregate
provider_name
Darkbloom
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-11-14T18:54:22.215Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
tau2-bench Verified Airline · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
500
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7633333999999999
accuracy_std_dev
0.022949248354575815
cost_per_task_usd
0.05667028799999999
output_tokens_per_task
53108.832
latency_seconds_per_task
1040.885294
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
8
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

79.25%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Alibaba
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
2
sample size
400
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
8
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
tau_bench_verified_airline
name
tau2-bench Verified Airline
record_kind
aggregate
provider_name
Alibaba
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T19:22:16.313Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
tau2-bench Verified Airline · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
400
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7924998750000001
accuracy_std_dev
0.02591377395535768
cost_per_task_usd
0.082230975
output_tokens_per_task
26801.935
latency_seconds_per_task
213.86449
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
29
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

77.041%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · CoreWeave
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
3
sample size
98
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
2
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
tau_bench_verified_airline
name
tau2-bench Verified Airline
record_kind
aggregate
provider_name
CoreWeave
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-08-29T17:18:52.408Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
tau2-bench Verified Airline · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
98
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7704079387755102
accuracy_std_dev
0.002986000000000044
cost_per_task_usd
0.1301587857142857
output_tokens_per_task
42669.47959183674
latency_seconds_per_task
225.4525
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
22
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

78.535%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Darkbloom
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
4
sample size
1980
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
10
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
Darkbloom
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-11-14T18:43:32.728Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
1980
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.7853533999999999
accuracy_std_dev
0.02150995189766822
cost_per_task_usd
0.020415863636363638
output_tokens_per_task
27143.04797979798
latency_seconds_per_task
455.44460656565656
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
10
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

83.203%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · CoreWeave
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
12
sample size
385
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
2
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
CoreWeave
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-08-29T18:29:13.946Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
385
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.8320348961038961
accuracy_std_dev
0.011673500000000003
cost_per_task_usd
0.02834737922077922
output_tokens_per_task
22533.716883116882
latency_seconds_per_task
116.34369090909091
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
27
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

73.218%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Parasail
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
15
sample size
346
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
7
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
tau_bench_verified_airline
name
tau2-bench Verified Airline
record_kind
aggregate
provider_name
Parasail
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T19:11:32.112Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
tau2-bench Verified Airline · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
346
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7321772312138729
accuracy_std_dev
0.08403648462496778
cost_per_task_usd
0.1502556156069364
output_tokens_per_task
112906.43641618497
latency_seconds_per_task
2018.2842369942198
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
28
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

82.849%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Alibaba
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
19
sample size
1584
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
8
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
Alibaba
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:08:08.538Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
1584
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.8284931249999998
accuracy_std_dev
0.006861031326220224
cost_per_task_usd
0.029469156565656567
output_tokens_per_task
22565.453914141413
latency_seconds_per_task
133.59733396464648
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
29
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

82.852%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · SiliconFlow
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
20
sample size
1386
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
7
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
SiliconFlow
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:10:31.145Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
1386
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.8285234285714286
accuracy_std_dev
0.008462147918097126
cost_per_task_usd
0.048541420634920646
output_tokens_per_task
26857.45165945166
latency_seconds_per_task
445.627215007215
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
34
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

81.578%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Venice
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
21
sample size
1386
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
7
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
Venice
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:09:44.821Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
1386
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.8157767142857143
accuracy_std_dev
0.008190962872306891
cost_per_task_usd
0.02609219480519481
output_tokens_per_task
23394.820346320346
latency_seconds_per_task
163.1665974025974
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
39
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

81.793%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Parasail
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
22
sample size
1385
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
7
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
Parasail
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:04:48.178Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
1385
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.8179302505415164
accuracy_std_dev
0.004936247761122178
cost_per_task_usd
0.02662889386281589
output_tokens_per_task
26506.814440433212
latency_seconds_per_task
314.14365198555953
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
36
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

68.095%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · AtlasCloud
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
22
sample size
350
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
7
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
tau_bench_verified_airline
name
tau2-bench Verified Airline
record_kind
aggregate
provider_name
AtlasCloud
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T19:46:05.991Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
tau2-bench Verified Airline · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
350
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.6809524285714286
accuracy_std_dev
0.09813702854661528
cost_per_task_usd
0.09356050285714285
output_tokens_per_task
21782.17142857143
latency_seconds_per_task
244.74557714285714
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
29
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

75.575%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · auto-routing
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
25
sample size
348
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
7
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
tau_bench_verified_airline
name
tau2-bench Verified Airline
record_kind
aggregate
provider_name
auto-routing
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T18:44:20.050Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
tau2-bench Verified Airline · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
348
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7557471896551724
accuracy_std_dev
0.051344765491603474
cost_per_task_usd
0.07958818965517242
output_tokens_per_task
47576.2183908046
latency_seconds_per_task
709.5093793103448
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
128
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

70.952%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Venice
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
25
sample size
350
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
7
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
tau_bench_verified_airline
name
tau2-bench Verified Airline
record_kind
aggregate
provider_name
Venice
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T19:22:37.818Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
tau2-bench Verified Airline · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
350
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7095237142857143
accuracy_std_dev
0.03614029729387676
cost_per_task_usd
0.06154300571428571
output_tokens_per_task
25975.411428571428
latency_seconds_per_task
247.46257428571428
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
38
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

78.836%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · AtlasCloud
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
26
sample size
1386
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
7
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
AtlasCloud
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:08:31.063Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
1386
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.7883595714285714
accuracy_std_dev
0.03112971003424745
cost_per_task_usd
0.03924310822510823
output_tokens_per_task
21702.200577200576
latency_seconds_per_task
198.69347186147186
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
30
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

82.408%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · DeepInfra
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
28
sample size
1531
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
8
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
DeepInfra
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:42:08.154Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
1531
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.8240802188112345
accuracy_std_dev
0.011135789059800642
cost_per_task_usd
0.02478115610711953
output_tokens_per_task
24668.269105160027
latency_seconds_per_task
443.1595382103201
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
71
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

OpenRouter Benchmarks

81.65%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · auto-routing
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
75
sample size
1386
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
7
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
auto-routing
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:05:36.303Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
1386
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.8164981428571428
accuracy_std_dev
0.013316748910714663
cost_per_task_usd
0.026554869408369406
output_tokens_per_task
25393.416305916307
latency_seconds_per_task
277.64969264069265
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
132
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3.5-35b-a3b-20260224
source model id
qwen/qwen3.5-35b-a3b-20260224
Original result ↗

Published prices

1–6 / 6

qwen/qwen3.5-35b-a3b ↗

$0.05

Inference provider
openRouter
Platform
OpenRouter
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions

qwen/qwen3.5-35b-a3b ↗

$0.14

Inference provider
openRouter
Platform
OpenRouter
Unit
input_tokens_per_million
Currency
USD
Billing conditions

qwen/qwen3.5-35b-a3b ↗

$1.00

Inference provider
openRouter
Platform
OpenRouter
Unit
output_tokens_per_million
Currency
USD
Billing conditions

Qwen/Qwen3.5-35B-A3B ↗

$0.15

Inference provider
siliconFlow
Platform
SiliconFlow
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
Billing label
Cached input tokens (per 1M)

Qwen/Qwen3.5-35B-A3B ↗

$0.24

Inference provider
siliconFlow
Platform
SiliconFlow
Unit
input_tokens_per_million
Currency
USD
Billing conditions
Billing label
Input tokens (per 1M)

Qwen/Qwen3.5-35B-A3B ↗

$1.80

Inference provider
siliconFlow
Platform
SiliconFlow
Unit
output_tokens_per_million
Currency
USD
Billing conditions
Billing label
Output tokens (per 1M)

Provider specifications & access

1–10 / 10

Qwen/Qwen3.5-35B-A3B ↗

source model id
azure-huggingface/qwen--qwen3.5-35b-a3b@1
hosting
foundry-managed-compute
platforms
  • Microsoft Foundry
license
apache-2.0
context length
262144
architecture
input_modalities
  • text
  • image
  • video
output_modalities
  • text
capabilities

    Qwen/Qwen3.5-35B-A3B ↗

    source model id
    azure-huggingface/qwen--qwen3.5-35b-a3b@2
    hosting
    foundry-managed-compute
    platforms
    • Microsoft Foundry
    license
    apache-2.0
    context length
    262144
    architecture
    input_modalities
    • text
    • image
    • video
    output_modalities
    • text
    capabilities

      Qwen/Qwen3.5-35B-A3B ↗

      source model id
      azure-huggingface/qwen--qwen3.5-35b-a3b@3
      hosting
      foundry-managed-compute
      platforms
      • Microsoft Foundry
      license
      apache-2.0
      context length
      262144
      architecture
      input_modalities
      • text
      • image
      • video
      output_modalities
      • text
      capabilities

        Qwen/Qwen3.5-35B-A3B ↗

        source model id
        azure-huggingface/qwen--qwen3.5-35b-a3b@4
        hosting
        foundry-managed-compute
        platforms
        • Microsoft Foundry
        license
        apache-2.0
        context length
        262144
        architecture
        input_modalities
        • text
        • image
        • video
        output_modalities
        • text
        capabilities

          Qwen/Qwen3.5-35B-A3B ↗

          source model id
          azure-huggingface/qwen--qwen3.5-35b-a3b@5
          hosting
          foundry-managed-compute
          platforms
          • Microsoft Foundry
          license
          apache-2.0
          context length
          262144
          architecture
          input_modalities
          • text
          • image
          • video
          output_modalities
          • text
          capabilities

            Qwen/Qwen3.5-35B-A3B ↗

            source model id
            azure-huggingface/qwen--qwen3.5-35b-a3b@6
            hosting
            foundry-managed-compute
            platforms
            • Microsoft Foundry
            license
            apache-2.0
            context length
            262144
            architecture
            input_modalities
            • text
            • image
            • video
            output_modalities
            • text
            capabilities

              Qwen/Qwen3.5-35B-A3B ↗

              source model id
              HuggingFace/qwen-qwen3.5-35b-a3b@1
              hosting
              foundry-managed-compute
              platforms
              • Microsoft Foundry
              license
              apache-2.0
              architecture
              input_modalities
              • text
              output_modalities
              • text
              capabilities

                Qwen/Qwen3.5-35B-A3B ↗

                source model id
                publishers/hf-qwen/models/qwen3.5-35b-a3b@001
                hosting
                vertex-self-deployed
                platforms
                • Google Cloud Vertex AI
                architecture
                input_modalities
                • image
                • text
                output_modalities
                • text
                capabilities
                • multimodal
                • visionInput

                Qwen3.5-35B-A3B ↗

                source model id
                publishers/qwen/models/qwen3-5@qwen3.5-35b-a3b
                hosting
                vertex-self-deployed
                platforms
                • Google Cloud Vertex AI
                architecture
                input_modalities
                • text
                output_modalities
                • text
                capabilities

                  Qwen3.5-35B-A3B ↗

                  source model id
                  Qwen/Qwen3.5-35B-A3B
                  hosting
                  siliconflow-hosted
                  platforms
                  • SiliconFlow
                  max output tokens
                  262144
                  architecture
                  input_modalities
                  • text
                  • image
                  output_modalities
                  • text
                  capabilities
                  • promptCaching
                  • reasoning
                  • structuredOutputs
                  • toolCalling

                  Safety assessments

                  1–1 / 1

                  Foundation Model Transparency Index 2025 ↗

                  26 / 100

                  source id
                  stanfordFMTI
                  canonical model id
                  qwen/qwen3.5-35b-a3b-20260224
                  taxonomy
                  transparency
                  hazard category
                  transparency
                  hazard name
                  Organization transparency
                  policy area
                  Alibaba
                  assessment type
                  organization-level disclosure
                  metric
                  disclosure_score
                  result label
                  26 / 100
                  sample size
                  100
                  updated at
                  2026-11-09T18:08:05.227Z

                  Source specification observations

                  1–10 / 10

                  azureFoundry ↗

                  field
                  context_length
                  value
                  262144
                  normalized value
                  262144
                  scope
                  source-model
                  source id
                  azureFoundry
                  confidence
                  1

                  azureFoundry ↗

                  field
                  context_length
                  value
                  262144
                  normalized value
                  262144
                  scope
                  source-model
                  source id
                  azureFoundry
                  confidence
                  1

                  azureFoundry ↗

                  field
                  context_length
                  value
                  262144
                  normalized value
                  262144
                  scope
                  source-model
                  source id
                  azureFoundry
                  confidence
                  1

                  azureFoundry ↗

                  field
                  context_length
                  value
                  262144
                  normalized value
                  262144
                  scope
                  source-model
                  source id
                  azureFoundry
                  confidence
                  1

                  azureFoundry ↗

                  field
                  context_length
                  value
                  262144
                  normalized value
                  262144
                  scope
                  source-model
                  source id
                  azureFoundry
                  confidence
                  1

                  azureFoundry ↗

                  field
                  context_length
                  value
                  262144
                  normalized value
                  262144
                  scope
                  source-model
                  source id
                  azureFoundry
                  confidence
                  1

                  openRouter ↗

                  field
                  context_length
                  value
                  262144
                  normalized value
                  262144
                  scope
                  source-model
                  source id
                  openRouter
                  confidence
                  1

                  siliconFlow ↗

                  field
                  context_length
                  value
                  262K
                  scope
                  source-model
                  source id
                  siliconFlow
                  confidence
                  1

                  openRouter ↗

                  field
                  max_output_tokens
                  value
                  81920
                  normalized value
                  81920
                  scope
                  source-model
                  source id
                  openRouter
                  confidence
                  1

                  siliconFlow ↗

                  field
                  max_output_tokens
                  value
                  262K
                  scope
                  source-model
                  source id
                  siliconFlow
                  confidence
                  1

                  Original sources

                  1–20 / 24

                  SiliconFlow · Model page ↗

                  Hugging Face · Model reference ↗

                  Google Cloud · Model page ↗

                  Google Cloud · Model page ↗

                  Microsoft Foundry · Model page ↗

                  Microsoft Foundry · Model page ↗

                  Microsoft Foundry · Model page ↗

                  Microsoft Foundry · Model page ↗

                  Microsoft Foundry · Model page ↗

                  Microsoft Foundry · Model page ↗

                  Microsoft Foundry · Model page ↗

                  OpenRouter · Model page ↗

                  SiliconFlow · Playground ↗

                  Google Cloud · Model documentation ↗

                  Microsoft Foundry · Model documentation ↗

                  SiliconFlow · API documentation ↗

                  Epoch AI · Catalog ↗

                  GitHub · Catalog ↗

                  Google Cloud · Catalog ↗

                  Hugging Face · Catalog ↗

                  Source links

                  1–9 / 9

                  Hugging Face model card

                  1–1 / 1

                  Qwen/Qwen3.5-35B-A3B ↗

                  downloads
                  1544044
                  likes
                  1559
                  pipeline tag
                  image-text-to-text
                  library name
                  transformers
                  License
                  apache-2.0

                  Explore more models

                  Models

                  Loading
                  EVIDENCE / COMPARISON

                  Benchmarks

                  Explore every benchmark

                  Loading

                  These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

                  Install TensorTrace