Mistral: Mistral Small 4

Mistral Small 4 is the next major release in the Mistral Small family, unifying the capabilities of several flagship Mistral models into a single system. It combines strong reasoning from Magistral, …

mistralai/mistral-small-2603

Mistral Small 4 is the next major release in the Mistral Small family, unifying the capabilities of several flagship Mistral models into a single system. It combines strong reasoning from Magistral, multimodal understanding from Pixtral, and agentic coding capabilities from Devstral, enabling one model to handle complex analysis, software development, and visual tasks within the same workflow.

Specifications & capabilities

Input
text, image
Output
text
Context window · tokens
262144
Published model ID
mistral-small-2603
Capabilities
batch, multimodal, promptCaching, reasoning, streaming, structuredOutputs, toolCalling, visionInput
Description source ↗

Evaluations

1–8 / 8

OpenRouter Benchmarks

75.281%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Mistral
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
5
sample size
1188
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
Mistral
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-09-17T02:17:49.284Z
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
1188
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.7528056666666667
accuracy_std_dev
0.004601995207395269
cost_per_task_usd
0.015185508417508417
output_tokens_per_task
24359.508417508416
latency_seconds_per_task
152.6224537037037
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
mistralai/mistral-small-2603
source model id
mistralai/mistral-small-2603
Original result ↗

OpenRouter Benchmarks

41.889%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · Mistral
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
5
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
Mistral
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-09-17T02:10:06.482Z
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
300
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.41888899999999996
accuracy_std_dev
0.014614444065147779
cost_per_task_usd
0.01463076
output_tokens_per_task
8814.213333333333
latency_seconds_per_task
76.54969333333332
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
mistralai/mistral-small-2603
source model id
mistralai/mistral-small-2603
Original result ↗

OpenRouter Benchmarks

75.589%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Venice
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
28
sample size
198
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
1
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
2026-09-01T12:59:55.618Z
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
198
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.755892
cost_per_task_usd
0.018171858585858586
output_tokens_per_task
24027.954545454544
latency_seconds_per_task
285.7600353535354
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
mistralai/mistral-small-2603
source model id
mistralai/mistral-small-2603
Original result ↗

OpenRouter Benchmarks

75.758%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · auto-routing
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
91
sample size
198
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
1
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
2026-09-01T13:30:26.223Z
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
198
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.7575759999999999
cost_per_task_usd
0.014556641414141414
output_tokens_per_task
23980.964646464647
latency_seconds_per_task
173.17991919191917
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
mistralai/mistral-small-2603
source model id
mistralai/mistral-small-2603
Original result ↗

OpenRouter Benchmarks

46%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · auto-routing
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
111
sample size
50
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
1
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
2026-09-01T13:02:23.335Z
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
50
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.46
cost_per_task_usd
0.011152899999999999
output_tokens_per_task
9087.9
latency_seconds_per_task
87.35746
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
mistralai/mistral-small-2603
source model id
mistralai/mistral-small-2603
Original result ↗

Published prices

1–7 / 7

mistral-small-2603 ↗

$0.15

Inference provider
mistral
Platform
Mistral
Unit
input_tokens_per_million
Currency
USD
Billing conditions
Billing label
input / 1M tokens

mistral-small-2603 ↗

$0.08

Inference provider
mistral
Platform
Mistral
Unit
input_tokens_per_million
Currency
USD
Billing conditions
inference_type
Batch API
Billing label
input / 1M tokens

mistral-small-2603 ↗

$0.60

Inference provider
mistral
Platform
Mistral
Unit
output_tokens_per_million
Currency
USD
Billing conditions
Billing label
output / 1M tokens

mistral-small-2603 ↗

$0.30

Inference provider
mistral
Platform
Mistral
Unit
output_tokens_per_million
Currency
USD
Billing conditions
inference_type
Batch API
Billing label
output / 1M tokens

Provider specifications & access

1–2 / 2

mistralai/Mistral-Small-4-119B-2603 ↗

source model id
HuggingFace/mistralai-mistral-small-4-119b-2603@1
hosting
foundry-managed-compute
platforms
  • Microsoft Foundry
license
apache-2.0
architecture
input_modalities
  • text
output_modalities
  • text
capabilities

    Mistral Small 4 ↗

    source model id
    mistral-small-2603
    hosting
    mistral-hosted
    platforms
    • Mistral
    context length
    256000
    architecture
    input_modalities
    • text
    • image
    output_modalities
    • text
    capabilities
    • batch
    • reasoning
    • streaming
    • structuredOutputs
    • toolCalling

    Safety assessments

    1–1 / 1

    Foundation Model Transparency Index 2025 ↗

    18 / 100

    source id
    stanfordFMTI
    canonical model id
    mistralai/mistral-small-2603
    taxonomy
    transparency
    hazard category
    transparency
    hazard name
    Organization transparency
    policy area
    Mistral
    assessment type
    organization-level disclosure
    metric
    disclosure_score
    result label
    18 / 100
    sample size
    100
    updated at
    2026-11-09T18:08:05.227Z

    Source specification observations

    1–3 / 3

    mistral ↗

    field
    context_length
    value
    256000
    normalized value
    256000
    scope
    source-model
    source id
    mistral
    confidence
    1

    openRouter ↗

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

    openRouter ↗

    field
    max_output_tokens
    value
    209715
    normalized value
    209715
    scope
    source-model
    source id
    openRouter
    confidence
    1

    Original sources

    1–15 / 15

    Mistral AI · Model reference ↗

    Hugging Face · Weights ↗

    Hugging Face · mistralai/Mistral-Small-4-119B-2603 ↗

    Microsoft Foundry · Model page ↗

    OpenRouter · Model page ↗

    Microsoft Foundry · Model documentation ↗

    Mistral AI · Model documentation ↗

    www.apache.org · Weights ↗

    Epoch AI · Catalog ↗

    GitHub · Catalog ↗

    Hugging Face · Catalog ↗

    Microsoft Foundry · Catalog ↗

    Mistral AI · Catalog ↗

    OpenRouter · Catalog ↗

    OpenRouter · Catalog ↗

    Source links

    1–9 / 9

    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