Mistral Small 4
Mistral AI · released
Mistral's low-cost workhorse: one Apache-2.0 model that covers fast chat, optional deep reasoning, coding and image input.
Best for: High-volume low-cost inference · Self-hosting · Document understanding
- Intelligence Index
- 11.3
- #38 of 38 · at Reasoning effort
- Price per 1M tokens
- $0.15 in $0.60 out
- Blended $0.26 at 3:1
- Context window
- 256K
- Output speed
- 160 tokens/s
- $0.02 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall0 · #38 of 38
Based on 1 of 3 benchmarks
- Coding5 · #37 of 39
Based on 2 of 5 benchmarks
- Agents & knowledge work0 · #39 of 40
Based on 1 of 6 benchmarks
- Reasoning & science0 · #39 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability17 · #37 of 39
Based on 2 of 2 benchmarks
- Vision0 · #31 of 31
Based on 1 of 2 benchmarks
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#38 of 38· Reasoning effort· Artificial Analysis
- Vals IndexNot tested
- Arena · TextNot tested
Coding
#35 of 38· Reasoning effort· Artificial Analysis
- SciCode38.8%
#36 of 38· Reasoning effort· Artificial Analysis
- Arena · WebDevNot tested
- Vibe Code BenchNot tested
- Code MigrationNot tested
Agents & knowledge work
- GDPval-AA0.0
#38 of 39· Reasoning effort· Artificial Analysis
- Arena · AgentNot tested
- Finance Agent v2Not tested
- Excel Modeling BenchmarkNot tested
- Legal Research BenchNot tested
- Tax Agent BenchNot tested
Reasoning & science
#39 of 39· Reasoning effort· Artificial Analysis
- CritPt0.3%
#35 of 39· Reasoning effort· Artificial Analysis
- ARC-AGI-2saturatingNot tested
- ARC-AGI-3Not tested
- GPQA Diamondsaturating76.9%
#25 of 27· Reasoning effort· Artificial Analysis
Knowledge & reliability
#34 of 39· Reasoning effort· Artificial Analysis
- AA-LCR49.7%
#39 of 39· Reasoning effort· Artificial Analysis
Vision
- MMMU-Pro56.8%
#27 of 27· Reasoning effort· Artificial Analysis
- Arena · VisionNot tested
Specifications
- Lab
- Mistral AI
- Released
- Mar 16, 2026
- Availability
- Generally available
- Context window
- 256,000 tokens
- Max output
- —
- Accepts
- Text, Images
- Knowledge cutoff
- Not published
- Weights
- Open
- Licence
- Apache 2.0
- Parameters
- 119B total · 6.5B active
- Reasoning control
- Effort none or high
- API model ID
mistral-small-2603
Pricing
- Input / 1M tokens
- $0.15
- Cached input / 1M
- $0.01
- Output / 1M tokens
- $0.60
- Blended (3:1)
- $0.26
- Measured cost / task
- $0.02
Batch processing is half price.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.