MoonshotAI: Kimi K3

Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at…

moonshotai/kimi-k3-20260715

Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at navigating large repositories, using tools, debugging, and iterating against images, logs, tests, and runtime feedback. Its architecture uses KDA and Attention Residuals for computational efficiency.

Specifications & capabilities

Input
text, image, video
Output
text
Context window · tokens
1048576
Published model ID
Kimi-K3
Capabilities
audioInput, multimodal, promptCaching, reasoning, structuredOutputs, toolCalling, videoInput, visionInput
Description source ↗

Evaluations

41–55 / 55

OpenRouter Benchmarks

73.306%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Sail Research
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
8
sample size
600
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
12
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
Sail Research
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-11-06T22:35:15.892Z
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
600
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7330556666666667
accuracy_std_dev
0.016468863254301695
cost_per_task_usd
0.1880071
output_tokens_per_task
9600.853333333333
latency_seconds_per_task
216.66543166666668
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
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

OpenRouter Benchmarks

72.713%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Fireworks
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
9
sample size
1097
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
22
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
Fireworks
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T12:24:25.742Z
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
1097
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7271345697356426
accuracy_std_dev
0.02666397891606509
cost_per_task_usd
0.4311616608933454
output_tokens_per_task
10876.747493163171
latency_seconds_per_task
251.08297538742025
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
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

Vercel AI Gateway

2.383%

source id
vercelAIGateway
leaderboard name
AI Gateway Text Tokens Share
dimension
popularity
metric
Tokens
rank
9
evaluated at
2026-11-11T00:00:00.000Z
source model id
Kimi K3
Original result ↗

OpenRouter Benchmarks

73.667%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Together
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
11
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
Together
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T13:04:39.070Z
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
0.7366666000000001
accuracy_std_dev
0.027039148574613085
cost_per_task_usd
0.29887413199999996
output_tokens_per_task
9417.472
latency_seconds_per_task
294.262764
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
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

OpenRouter Benchmarks

72.061%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · DigitalOcean
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
12
sample size
550
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
11
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
DigitalOcean
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T12:33:51.242Z
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
550
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7206058181818182
accuracy_std_dev
0.06497575192305928
cost_per_task_usd
0.2521624945454546
output_tokens_per_task
10421.192727272728
latency_seconds_per_task
431.8623145454546
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
20
benchmark_evaluation
true
openrouter_permaslug
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

OpenRouter Benchmarks

73.778%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Parasail
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
13
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
Parasail
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T11:28:52.482Z
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
0.7377776666666667
accuracy_std_dev
0.020366893981273733
cost_per_task_usd
0.3511924911111111
output_tokens_per_task
9122.282222222222
latency_seconds_per_task
195.77759333333333
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
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

OpenRouter Benchmarks

73.494%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Phala
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
13
sample size
498
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
Phala
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T12:42:56.001Z
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
498
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7349397911646587
accuracy_std_dev
0.047428285455095254
cost_per_task_usd
0.2969604819277108
output_tokens_per_task
11351.530120481928
latency_seconds_per_task
386.5659397590361
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
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

OpenRouter Benchmarks

92.02%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · auto-routing
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
14
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
auto-routing
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T08:21:51.280Z
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
0.9202018
accuracy_std_dev
0.01007849159150319
cost_per_task_usd
0.13160375303030303
output_tokens_per_task
10658.989393939393
latency_seconds_per_task
237.24880202020202
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
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

OpenRouter Benchmarks

74.242%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Alibaba
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
16
sample size
550
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
11
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
2026-10-31T12:43:33.294Z
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
550
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.7424241818181818
accuracy_std_dev
0.03530779679235989
cost_per_task_usd
0.36784247818181814
output_tokens_per_task
9773.532727272728
latency_seconds_per_task
381.85354363636367
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
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

OpenRouter Benchmarks

73.333%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Amazon Bedrock
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
16
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
Amazon Bedrock
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T12:51:11.835Z
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
0.7333334
accuracy_std_dev
0.03451234883690184
cost_per_task_usd
0.30572972400000004
output_tokens_per_task
10813.856
latency_seconds_per_task
282.521796
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
31
benchmark_evaluation
true
openrouter_permaslug
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

OpenRouter Benchmarks

74.133%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · DeepInfra
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
26
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
DeepInfra
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T12:27:47.241Z
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
0.7413333
accuracy_std_dev
0.03095523075685272
cost_per_task_usd
0.40647999399999996
output_tokens_per_task
9896.416
latency_seconds_per_task
588.53878
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
63
benchmark_evaluation
true
openrouter_permaslug
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

OpenRouter Benchmarks

73.852%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · auto-routing
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
40
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
auto-routing
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-31T11:38:53.393Z
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
0.7385184444444445
accuracy_std_dev
0.028765060349555692
cost_per_task_usd
0.24696524444444445
output_tokens_per_task
10562.944444444445
latency_seconds_per_task
265.21315111111113
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
moonshotai/kimi-k3-20260715
source model id
moonshotai/kimi-k3-20260715
Original result ↗

Published prices

1–20 / 93

moonshotai/kimi-k3 ↗

$0.30

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

moonshotai/kimi-k3 ↗

$3.00

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

moonshotai/kimi-k3 ↗

$15.00

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

moonshotai/kimi-k3 ↗

$0.18

Inference provider
openRouter
Platform
OpenRouter
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions

moonshotai/kimi-k3 ↗

$0.19

Inference provider
openRouter
Platform
OpenRouter
Unit
input_tokens_per_million
Currency
USD
Billing conditions

moonshotai/kimi-k3 ↗

$16.00

Inference provider
openRouter
Platform
OpenRouter
Unit
output_tokens_per_million
Currency
USD
Billing conditions

moonshotai/kimi-k3:batch ↗

$0.23

Inference provider
openRouter
Platform
OpenRouter
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
mode
batch

moonshotai/kimi-k3:batch ↗

$2.28

Inference provider
openRouter
Platform
OpenRouter
Unit
input_tokens_per_million
Currency
USD
Billing conditions
mode
batch

moonshotai/kimi-k3:batch ↗

$11.40

Inference provider
openRouter
Platform
OpenRouter
Unit
output_tokens_per_million
Currency
USD
Billing conditions
mode
batch

~moonshotai/kimi-latest ↗

$0.18

Inference provider
openRouter
Platform
OpenRouter
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
alias_target
name
MoonshotAI: Kimi K3
slug
moonshotai/kimi-k3
routing_alias
true

~moonshotai/kimi-latest ↗

$0.18

Inference provider
openRouter
Platform
OpenRouter
Unit
input_tokens_per_million
Currency
USD
Billing conditions
alias_target
name
MoonshotAI: Kimi K3
slug
moonshotai/kimi-k3
routing_alias
true

~moonshotai/kimi-latest ↗

$11.01

Inference provider
openRouter
Platform
OpenRouter
Unit
output_tokens_per_million
Currency
USD
Billing conditions
alias_target
name
MoonshotAI: Kimi K3
slug
moonshotai/kimi-k3
routing_alias
true

moonshotai/Kimi-K3 ↗

$0.27

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

moonshotai/Kimi-K3 ↗

$2.70

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

moonshotai/Kimi-K3 ↗

$13.50

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

moonshotai/kimi-k3 ↗

$0.30

Inference provider
vercelAIGatewayCatalog
Platform
alibaba
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
endpoint
alibaba | moonshotai/kimi-k3
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Cached input / 1M
Pricing via
Vercel AI Gateway

moonshotai/kimi-k3 ↗

$3.00

Inference provider
vercelAIGatewayCatalog
Platform
alibaba
Unit
input_tokens_per_million
Currency
USD
Billing conditions
endpoint
alibaba | moonshotai/kimi-k3
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Input tokens / 1M
Pricing via
Vercel AI Gateway

moonshotai/kimi-k3 ↗

$15.00

Inference provider
vercelAIGatewayCatalog
Platform
alibaba
Unit
output_tokens_per_million
Currency
USD
Billing conditions
endpoint
alibaba | moonshotai/kimi-k3
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Output tokens / 1M
Pricing via
Vercel AI Gateway

moonshotai/kimi-k3 ↗

$0.30

Inference provider
vercelAIGatewayCatalog
Platform
baseten
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
endpoint
baseten | moonshotai/kimi-k3
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Cached input / 1M
Pricing via
Vercel AI Gateway

moonshotai/kimi-k3 ↗

$0.30

Inference provider
vercelAIGatewayCatalog
Platform
baseten
Unit
cache_read_tokens_per_million
Currency
USD
Billing conditions
region
scope
zone
geo_region
us
endpoint
baseten | moonshotai/kimi-k3
model_type
language
billing_scope
via Vercel AI Gateway
Billing label
Cached input / 1M
Pricing via
Vercel AI Gateway

Provider specifications & access

1–5 / 5

moonshot ai ↗

source model id
Kimi K3
hosting
copilot-subscription
platforms
  • GitHub Copilot
architecture
license
CC-BY-4.0
attribution
GitHub and github/docs contributors
plans
max
true
pro
true
business
true
pro_plus
true
enterprise
true
clients
vs
true
cli
true
xcode
true
dotcom
true
vscode
true
eclipse
true
jetbrains
true
access scope
Published plan/client flags; account settings, organization policy and additional access conditions apply
availability
supported
release status
GA
billing scope
Copilot subscription usage; not an unconditional inference API rate
billing rates
  • unit
    tokens_per_million
    input
    3
    output
    15
    currency
    USD
    original
    input
    $3.00
    model
    Kimi K3
    output
    $15.00
    category
    Powerful
    provider
    moonshot_ai
    cached_input
    $0.30
    release_status
    GA
    cached_input
    0.3
    footnote_restrictions
    false
    requires_subscription
    true
comparison
name
Kimi K3
excels_at
Multi-step agent tasks across large codebases
task_area
Agentic coding and long-context work
further_reading
[Kimi K3 model card](https://huggingface.co/moonshotai/Kimi-K3)

Kimi K3 ↗

source model id
moonshotai.kimi-k3
hosting
bedrock-managed
platforms
  • Amazon Bedrock
architecture
input_modalities
  • text
  • image
output_modalities
  • text
capabilities

    Kimi K3 ↗

    source model id
    moonshotai/kimi-k3
    hosting
    third-party
    platforms
    • moonshot
    context length
    1048576
    max output tokens
    1048576
    architecture
    input_modalities
    • text
    • audio
    output_modalities
    • text
    capabilities
    • toolCalling
    • reasoning

    Kimi K3 ↗

    source model id
    moonshotai/kimi-k3
    hosting
    vercel-gateway
    platforms
    • Vercel AI Gateway
    context length
    1000000
    max output tokens
    131072
    architecture
    input_modalities
    • text
    • image
    • pdf
    • video
    output_modalities
    • text
    capabilities
    • promptCaching
    • reasoning
    • structuredOutputs
    • toolCalling
    • videoInput
    • visionInput
    regions
    • us

    Kimi-K3 ↗

    source model id
    moonshotai/Kimi-K3
    hosting
    siliconflow-hosted
    platforms
    • SiliconFlow
    max output tokens
    262144
    architecture
    input_modalities
    • text
    • image
    output_modalities
    • text
    capabilities
    • promptCaching
    • structuredOutputs
    • toolCalling

    Source specification observations

    1–12 / 12

    cloudflareWorkersAI ↗

    field
    context_length
    value
    1048576
    normalized value
    1048576
    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

    openRouter ↗

    field
    context_length
    value
    1048576
    normalized value
    1048576
    scope
    source-model
    source id
    openRouter
    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
    1000000
    normalized value
    1000000
    scope
    source-model
    source id
    vercelAIGatewayCatalog
    confidence
    1

    cloudflareWorkersAI ↗

    field
    max_output_tokens
    value
    1048576
    normalized value
    1048576
    scope
    source-model
    source id
    cloudflareWorkersAI
    confidence
    1

    openRouter ↗

    field
    max_output_tokens
    value
    943718
    normalized value
    943718
    scope
    source-model
    source id
    openRouter
    confidence
    1

    openRouter ↗

    field
    max_output_tokens
    value
    16384
    normalized value
    16384
    scope
    source-model
    source id
    openRouter
    confidence
    1

    openRouter ↗

    field
    max_output_tokens
    value
    943718
    normalized value
    943718
    scope
    source-model
    source id
    openRouter
    confidence
    1

    siliconFlow ↗

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

    vercelAIGatewayCatalog ↗

    field
    max_output_tokens
    value
    131072
    normalized value
    131072
    scope
    source-model
    source id
    vercelAIGatewayCatalog
    confidence
    1

    Original sources

    1–20 / 21

    SiliconFlow · Model page ↗

    Hugging Face · Model reference ↗

    OpenRouter · Model page ↗

    OpenRouter · Model page ↗

    SiliconFlow · Playground ↗

    Vercel AI Gateway · Model page ↗

    GitHub · Documentation ↗

    SiliconFlow · API documentation ↗

    GitHub · Source snapshot ↗

    Amazon Web Services · Model page ↗

    Amazon Bedrock · Model documentation ↗

    Amazon Web Services · Catalog ↗

    Cloudflare AI · Catalog ↗

    Epoch AI · Catalog ↗

    Hugging Face · Catalog ↗

    Hugging Face · Catalog ↗

    OpenRouter · Catalog ↗

    OpenRouter · Catalog ↗

    SiliconFlow · Catalog ↗

    Vercel AI Gateway · Catalog ↗

    Source links

    1–12 / 12

    Hugging Face model card

    1–1 / 1

    moonshotai/Kimi-K3 ↗

    downloads
    862111
    likes
    11871
    pipeline tag
    image-text-to-text
    library name
    transformers
    License
    other

    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