DeepSeek: DeepSeek V4 Pro 0423

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reas…

deepseek/deepseek-v4-pro-20260423

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding, and long-horizon agent workflows, with strong performance across knowledge, math, and software engineering benchmarks.

Built on the same architecture as DeepSeek V4 Flash, it introduces a hybrid attention system for efficient long-context processing. Reasoning efforts high and xhigh are supported; xhigh maps to max reasoning. It is well suited for complex workloads such as full-codebase analysis, multi-step automation, and large-scale information synthesis, where both capability and efficiency are critical

Specifications & capabilities

Input
text
Output
text
Context window · tokens
1048576
Knowledge cutoff
2025-05
Published model ID
DeepSeek-V4-Pro
Capabilities
audioInput, multimodal, promptCaching, reasoning, structuredOutputs, toolCalling
Description source ↗

Evaluations

1–20 / 67

OpenRouter Benchmarks

91.688%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Cloudflare
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
1
sample size
1185
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
6
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
Cloudflare
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:24:40.574Z
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
1185
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.916877387341772
accuracy_std_dev
0.005308941032719149
cost_per_task_usd
0.10198255443037975
output_tokens_per_task
39607.979746835445
latency_seconds_per_task
806.6257974683544
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
18
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

77.067%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · NextBit
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
1
sample size
250
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
5
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
NextBit
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-03T16:07:05.245Z
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
250
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7706666
accuracy_std_dev
0.0204828947622156
cost_per_task_usd
0.10916058
output_tokens_per_task
11359.708
latency_seconds_per_task
258.12842
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
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenCompass LLM

97.1%

source id
openCompass
leaderboard name
Knowledge_Science
dimension
general
metric
Knowledge_Science
rank
2
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
KnowledgeTable
catalog
closed
snapshot
month
26-07
fileName
assets/llm-rank/llm-data-v2.26-07.20260824.json
updateTime
2026-08-24
columnFileName
assets/llm-rank/llm-column-v2.26-07.20260824.json
metric_kind
Knowledge_Science
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
1-data.json
source_revision
117c5effa3066abf39f3e8f61109707b09f0cd95223246c07acd6289f4f7fcd2
metric_population
15
benchmark_evaluation
true
source model id
DeepSeek-V4-Pro
Original result ↗

OpenCompass LLM

95%

source id
openCompass
leaderboard name
Knowledge_Humanity
dimension
general
metric
Knowledge_Humanity
rank
2
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
KnowledgeTable
catalog
closed
snapshot
month
26-07
fileName
assets/llm-rank/llm-data-v2.26-07.20260824.json
updateTime
2026-08-24
columnFileName
assets/llm-rank/llm-column-v2.26-07.20260824.json
metric_kind
Knowledge_Humanity
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
1-data.json
source_revision
117c5effa3066abf39f3e8f61109707b09f0cd95223246c07acd6289f4f7fcd2
metric_population
15
benchmark_evaluation
true
source model id
DeepSeek-V4-Pro
Original result ↗

OpenCompass LLM

92.5%

source id
openCompass
leaderboard name
Knowledge_CommonSense
dimension
general
metric
Knowledge_CommonSense
rank
2
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
KnowledgeTable
catalog
closed
snapshot
month
26-07
fileName
assets/llm-rank/llm-data-v2.26-07.20260824.json
updateTime
2026-08-24
columnFileName
assets/llm-rank/llm-column-v2.26-07.20260824.json
metric_kind
Knowledge_CommonSense
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
1-data.json
source_revision
117c5effa3066abf39f3e8f61109707b09f0cd95223246c07acd6289f4f7fcd2
metric_population
15
benchmark_evaluation
true
source model id
DeepSeek-V4-Pro
Original result ↗

OpenRouter Benchmarks

78.597%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · BaseTen
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
2
sample size
95
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
BaseTen
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-08-29T17:15:26.910Z
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
95
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7859650631578948
accuracy_std_dev
0.03135199999999999
cost_per_task_usd
0.09807105263157893
output_tokens_per_task
8018.557894736842
latency_seconds_per_task
69.95393684210526
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
14
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

76.889%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Cloudflare
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
2
sample size
300
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
6
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
Cloudflare
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T20:34:04.127Z
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
300
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7688888333333334
accuracy_std_dev
0.013698652724710655
cost_per_task_usd
0.15528754666666666
output_tokens_per_task
20104.073333333334
latency_seconds_per_task
428.0695
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
15
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

76.556%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · SiliconFlow
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
2
sample size
300
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
6
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
SiliconFlow
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T21:36:17.452Z
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
300
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7655555
accuracy_std_dev
0.04258615812972569
cost_per_task_usd
0.10541816000000001
output_tokens_per_task
11441.496666666666
latency_seconds_per_task
274.1540233333333
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
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

76.519%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · GMICloud
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
2
sample size
450
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
9
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
GMICloud
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-11-10T08:00:49.752Z
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
450
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7651852222222222
accuracy_std_dev
0.027893614278292685
cost_per_task_usd
0.09080310222222222
output_tokens_per_task
18878.693333333333
latency_seconds_per_task
365.59188666666665
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
23
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

76.476%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Baidu
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
2
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
Baidu
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T20:06:41.224Z
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
0.7647619999999999
accuracy_std_dev
0.026116980863579387
cost_per_task_usd
0.04505181428571428
output_tokens_per_task
13672.442857142858
latency_seconds_per_task
210.43656857142855
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
11
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

75%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Ionstream
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
2
sample size
100
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
Ionstream
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-08-29T17:19:42.622Z
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
100
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.75
accuracy_std_dev
0.030000000000000027
cost_per_task_usd
0.08026519
output_tokens_per_task
11165.01
latency_seconds_per_task
649.8940799999999
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
5
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

90.38%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Novita
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
3
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
Novita
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:10:36.257Z
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
0.9037997142857144
accuracy_std_dev
0.016282944378479363
cost_per_task_usd
0.12103858297258295
output_tokens_per_task
37524.782106782106
latency_seconds_per_task
424.14833405483404
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
59
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

88.817%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Crusoe
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
3
sample size
386
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
Crusoe
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T16:08:08.539Z
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
386
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.8881691813471503
accuracy_std_dev
0.0030120000000000147
cost_per_task_usd
0.1504010103626943
output_tokens_per_task
42839.41709844559
latency_seconds_per_task
651.1954274611398
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
15
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenCompass LLM

92.5%

source id
openCompass
leaderboard name
Knowledge
dimension
general
metric
Knowledge
rank
4
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
OverallTable
catalog
closed
snapshot
month
26-07
fileName
assets/llm-rank/llm-data-v2.26-07.20260824.json
updateTime
2026-08-24
columnFileName
assets/llm-rank/llm-column-v2.26-07.20260824.json
metric_kind
Knowledge
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
1-data.json
source_revision
117c5effa3066abf39f3e8f61109707b09f0cd95223246c07acd6289f4f7fcd2
metric_population
15
benchmark_evaluation
true
source model id
DeepSeek-V4-Pro
Original result ↗

OpenCompass LLM

92.5%

source id
openCompass
leaderboard name
Average
dimension
general
metric
Average
rank
4
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
KnowledgeTable
catalog
closed
snapshot
month
26-07
fileName
assets/llm-rank/llm-data-v2.26-07.20260824.json
updateTime
2026-08-24
columnFileName
assets/llm-rank/llm-column-v2.26-07.20260824.json
metric_kind
Average
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
1-data.json
source_revision
117c5effa3066abf39f3e8f61109707b09f0cd95223246c07acd6289f4f7fcd2
metric_population
15
benchmark_evaluation
true
source model id
DeepSeek-V4-Pro
Original result ↗

OpenRouter Benchmarks

88.754%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · NextBit
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
4
sample size
987
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
5
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
NextBit
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-03T16:44:56.199Z
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
987
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.887537924012158
accuracy_std_dev
0.020250680116973837
cost_per_task_usd
0.12033997061803446
output_tokens_per_task
34357.5663627153
latency_seconds_per_task
580.7554113475178
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
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

83.165%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Ionstream
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
4
sample size
396
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
Ionstream
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-08-29T19:26:41.546Z
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
396
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.8316495
accuracy_std_dev
0.055555500000000035
cost_per_task_usd
0.04943669696969697
output_tokens_per_task
25885.7702020202
latency_seconds_per_task
1128.190707070707
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
5
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

76.952%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Reka
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
4
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
Reka
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-23T07:26:58.509Z
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
0.7695237142857143
accuracy_std_dev
0.02386662296246901
cost_per_task_usd
0.11461242285714283
output_tokens_per_task
11699.722857142857
latency_seconds_per_task
472.9290742857143
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
12
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenRouter Benchmarks

75.667%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · StreamLake
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
4
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
StreamLake
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T20:10:47.431Z
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
0.756666625
accuracy_std_dev
0.03366510310001107
cost_per_task_usd
0.01679233
output_tokens_per_task
10730.4225
latency_seconds_per_task
291.9460025
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
25
benchmark_evaluation
true
openrouter_permaslug
deepseek/deepseek-v4-pro-20260423
source model id
deepseek/deepseek-v4-pro-20260423
Original result ↗

OpenCompass LLM

91%

source id
openCompass
leaderboard name
Comprehensive_Coding_Ability
dimension
general
metric
Comprehensive_Coding_Ability
rank
5
aggregation scope
status
published
direction
higherIsBetter
conditions
suite
CodeTable
catalog
closed
snapshot
month
26-07
fileName
assets/llm-rank/llm-data-v2.26-07.20260824.json
updateTime
2026-08-24
columnFileName
assets/llm-rank/llm-column-v2.26-07.20260824.json
metric_kind
Comprehensive_Coding_Ability
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
1-data.json
source_revision
117c5effa3066abf39f3e8f61109707b09f0cd95223246c07acd6289f4f7fcd2
metric_population
15
benchmark_evaluation
true
source model id
DeepSeek-V4-Pro
Original result ↗

Published prices

1–20 / 36

deepseek/deepseek-v4-pro ↗

$0.15

Inference provider
cloudflareWorkersAI
Platform
Fireworks AI
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
Billing label
Cached input tokens (per 1M)
Pricing via
Cloudflare AI Gateway

deepseek/deepseek-v4-pro ↗

$1.74

Inference provider
cloudflareWorkersAI
Platform
Fireworks AI
Unit
input_tokens_per_million
Currency
USD
Billing conditions
Billing label
Input tokens (per 1M)
Pricing via
Cloudflare AI Gateway

deepseek/deepseek-v4-pro ↗

$3.48

Inference provider
cloudflareWorkersAI
Platform
Fireworks AI
Unit
output_tokens_per_million
Currency
USD
Billing conditions
Billing label
Output tokens (per 1M)
Pricing via
Cloudflare AI Gateway

deepseek/deepseek-v4-pro ↗

$0.07

Inference provider
openRouter
Platform
OpenRouter
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions

deepseek-ai/DeepSeek-V4-Pro ↗

$0.14

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

deepseek-ai/DeepSeek-V4-Pro ↗

$1.50

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

deepseek-ai/DeepSeek-V4-Pro ↗

$3.14

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

deepseek/deepseek-v4-pro ↗

$0.20

Inference provider
vercelAIGatewayCatalog
Platform
alibaba
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
endpoint
alibaba | deepseek/deepseek-v4-pro
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Cached input / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$2.40

Inference provider
vercelAIGatewayCatalog
Platform
alibaba
Unit
input_tokens_per_million
Currency
USD
Billing conditions
endpoint
alibaba | deepseek/deepseek-v4-pro
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Input tokens / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$4.80

Inference provider
vercelAIGatewayCatalog
Platform
alibaba
Unit
output_tokens_per_million
Currency
USD
Billing conditions
endpoint
alibaba | deepseek/deepseek-v4-pro
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Output tokens / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$0.15

Inference provider
vercelAIGatewayCatalog
Platform
azure
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
endpoint
azure | deepseek/deepseek-v4-pro
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Cached input / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$1.74

Inference provider
vercelAIGatewayCatalog
Platform
azure
Unit
input_tokens_per_million
Currency
USD
Billing conditions
endpoint
azure | deepseek/deepseek-v4-pro
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Input tokens / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$3.48

Inference provider
vercelAIGatewayCatalog
Platform
azure
Unit
output_tokens_per_million
Currency
USD
Billing conditions
endpoint
azure | deepseek/deepseek-v4-pro
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Output tokens / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$0.13

Inference provider
vercelAIGatewayCatalog
Platform
baseten
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
endpoint
baseten | deepseek/deepseek-v4-pro
model_type
language
quantization
fp4
billing_scope
via Vercel AI Gateway
Billing label
Cached input / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$0.13

Inference provider
vercelAIGatewayCatalog
Platform
baseten
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
region
scope
zone
geo_region
us
endpoint
baseten | deepseek/deepseek-v4-pro
model_type
language
quantization
fp4
billing_scope
via Vercel AI Gateway
Billing label
Cached input / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$1.32

Inference provider
vercelAIGatewayCatalog
Platform
baseten
Unit
input_tokens_per_million
Currency
USD
Billing conditions
endpoint
baseten | deepseek/deepseek-v4-pro
model_type
language
quantization
fp4
billing_scope
via Vercel AI Gateway
Billing label
Input tokens / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$1.32

Inference provider
vercelAIGatewayCatalog
Platform
baseten
Unit
input_tokens_per_million
Currency
USD
Billing conditions
region
scope
zone
geo_region
us
endpoint
baseten | deepseek/deepseek-v4-pro
model_type
language
quantization
fp4
billing_scope
via Vercel AI Gateway
Billing label
Input tokens / 1M
Pricing via
Vercel AI Gateway

deepseek/deepseek-v4-pro ↗

$3.96

Inference provider
vercelAIGatewayCatalog
Platform
baseten
Unit
output_tokens_per_million
Currency
USD
Billing conditions
endpoint
baseten | deepseek/deepseek-v4-pro
model_type
language
quantization
fp4
billing_scope
via Vercel AI Gateway
Billing label
Output tokens / 1M
Pricing via
Vercel AI Gateway

Provider specifications & access

1–4 / 4

DeepSeek-V4-Pro ↗

source model id
deepseek-ai/DeepSeek-V4-Pro
hosting
siliconflow-hosted
platforms
  • SiliconFlow
max output tokens
393216
architecture
input_modalities
  • text
output_modalities
  • text
capabilities
  • promptCaching
  • reasoning
  • structuredOutputs
  • toolCalling

DeepSeek V4 Pro ↗

source model id
deepseek/deepseek-v4-pro
hosting
third-party
platforms
  • Fireworks AI
context length
131072
architecture
input_modalities
  • text
  • audio
output_modalities
  • text
capabilities

    DeepSeek V4 Pro ↗

    source model id
    deepseek/deepseek-v4-pro
    hosting
    vercel-gateway
    platforms
    • Vercel AI Gateway
    context length
    1048600
    max output tokens
    1048600
    architecture
    input_modalities
    • text
    output_modalities
    • text
    capabilities
    • promptCaching
    • reasoning
    • structuredOutputs
    • toolCalling
    regions
    • us

    deepseek-ai/DeepSeek-V4-Pro ↗

    source model id
    publishers/deepseek-ai/models/deepseek-v4@deepseek-v4-pro
    hosting
    vertex-self-deployed
    platforms
    • Google Cloud Vertex AI
    architecture
    input_modalities
    • text
    output_modalities
    • text
    capabilities

      Safety assessments

      1–1 / 1

      Foundation Model Transparency Index 2025 ↗

      32 / 100

      source id
      stanfordFMTI
      canonical model id
      deepseek/deepseek-v4-pro-20260423
      taxonomy
      transparency
      hazard category
      transparency
      hazard name
      Organization transparency
      policy area
      DeepSeek
      assessment type
      organization-level disclosure
      metric
      disclosure_score
      result label
      32 / 100
      sample size
      100
      updated at
      2026-11-09T18:08:05.227Z

      Source specification observations

      1–8 / 8

      cloudflareWorkersAI ↗

      field
      context_length
      value
      131072
      normalized value
      131072
      scope
      source-model
      source id
      cloudflareWorkersAI
      confidence
      1

      openRouter ↗

      field
      context_length
      value
      1048576
      normalized value
      1048576
      scope
      source-model
      source id
      openRouter
      confidence
      1

      siliconFlow ↗

      field
      context_length
      value
      1049K
      scope
      source-model
      source id
      siliconFlow
      confidence
      1

      vercelAIGatewayCatalog ↗

      field
      context_length
      value
      1048600
      normalized value
      1048600
      scope
      source-model
      source id
      vercelAIGatewayCatalog
      confidence
      1

      vercelAIGatewayCatalog ↗

      field
      knowledge_cutoff
      reported field
      knowledge_cutoff
      value
      2025-05
      normalized value
      2025-05
      precision
      month
      scope
      source-model
      source id
      vercelAIGatewayCatalog
      confidence
      1

      openRouter ↗

      field
      max_output_tokens
      value
      384000
      normalized value
      384000
      scope
      source-model
      source id
      openRouter
      confidence
      1

      siliconFlow ↗

      field
      max_output_tokens
      value
      393K
      scope
      source-model
      source id
      siliconFlow
      confidence
      1

      vercelAIGatewayCatalog ↗

      field
      max_output_tokens
      value
      1048600
      normalized value
      1048600
      scope
      source-model
      source id
      vercelAIGatewayCatalog
      confidence
      1

      Original sources

      1–19 / 19

      SiliconFlow · Model page ↗

      Hugging Face · Model repository ↗

      Google Cloud · Model page ↗

      OpenRouter · Model page ↗

      SiliconFlow · Playground ↗

      Vercel AI Gateway · Model page ↗

      Google Cloud · Model documentation ↗

      SiliconFlow · API documentation ↗

      GitHub · Source snapshot ↗

      Cloudflare AI · Catalog ↗

      GitHub · Catalog ↗

      Google Cloud · Catalog ↗

      Hugging Face · Catalog ↗

      Hugging Face · Catalog ↗

      Kaggle · Catalog ↗

      OpenRouter · Catalog ↗

      OpenRouter · Catalog ↗

      SiliconFlow · Catalog ↗

      Vercel AI Gateway · Catalog ↗

      Source links

      1–12 / 12

      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