vs
2:6
overall capability wins (2 ties across 10 scored). Mistral Small 4 leads overall.
- Reasoning0:5
- Coding1:1
- Tool use1:0
- Composite indices0:2
- Pricing0:3
- Speed0:2
Verdict
Mistral Small 4 edges reasoning & knowledge; Mistral Small 4 is ~3.0x cheaper on a blended token basis than Devstral 2.
- Mistral Small 4 edges reasoning & knowledge
- Mistral Small 4 is ~3.0x cheaper on a blended token basis than Devstral 2
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Devstral 2
- $0.90
- Mistral Small 4
- $0.30
Mistral Small 4 is about 3.0x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Mistral
Mistral
Family
Mistral
Mistral
License
Apache 2.0
Apache 2.0
Open weights
Yes
Yes
Release
Dec 1, 2025
Mar 1, 2026
Knowledge cutoff
-
-
API / provider
Mistral
Mistral
Modalities
text → text
text, image → text
Specs
Context window
256K
256K
Max output
64K
64K
Parameters
123B
—
Pricing
Input $/1M
$0.40
$0.15
Output $/1M
$2
$0.60
Blended $/1M (3∶1)
$0.80
$0.26
Speed
tok/s
85
140
TTFT (s)
0.35
0.2
Reasoning
MMLU-Pro
76.2
—
GPQA Diamond
59.4
76.9
Humanity's Last Exam
3.6
9.9
AIME 2025
36.7
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
20.8
21.7
AA-LCR v1.1
32.3
49.7
CritPt
—
0.3
MMMU-Pro
—
56.8
IFBench
38.1
48.2
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
44.8
—
Terminal-Bench 2.1
30.3
21.0
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
32.8
38.8
CursorBench
—
—
SWE-Rebench
42.0
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
1,194
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | Devstral 2 | Mistral Small 4 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Mistral | Mistral | — |
| Family | Mistral | Mistral | — |
| License | Apache 2.0 | Apache 2.0 | — |
| Open weights | Yes | Yes | — |
| Release | Dec 1, 2025 | Mar 1, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Mistral | Mistral | — |
| Modalities | text → text | text, image → text | — |
| Specs | |||
| Context window | 256K | 256K | tie |
| Max output | 64K | 64K | tie |
| Parameters | 123B | — | — |
| Pricing | |||
| Input $/1M | $0.40 | $0.15 | B +$0.25 |
| Output $/1M | $2 | $0.60 | B +$1.40 |
| Blended $/1M (3∶1) | $0.80 | $0.26 | B +$0.54 |
| Speed | |||
| tok/s | 85 | 140 | B +55 |
| TTFT (s) | 0.35 | 0.2 | B +0.15 |
| Reasoning | |||
| MMLU-Pro | 76.2 | — | — |
| GPQA Diamond | 59.4 | 76.9 | B +17.5 pts |
| Humanity's Last Exam | 3.6 | 9.9 | B +6.3 pts |
| AIME 2025 | 36.7 | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 20.8 | 21.7 | B +0.9 pts |
| AA-LCR v1.1 | 32.3 | 49.7 | B +17.4 pts |
| CritPt | — | 0.3 | — |
| MMMU-Pro | — | 56.8 | — |
| IFBench | 38.1 | 48.2 | B +10.1 pts |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 44.8 | — | — |
| Terminal-Bench 2.1 | 30.3 | 21.0 | A +9.3 pts |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 32.8 | 38.8 | B +6.0 pts |
| CursorBench | — | — | — |
| SWE-Rebench | 42.0 | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | 1,194 | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
Devstral 2
- Context
- 256K / 64K out
- Parameters
- 123B
- Price
- $0.40 / $2
- Speed
- 85 tok/s · 0.35s TTFT
- Modalities
- text → text
- License
- Apache 2.0
Open-weight agentic coding model for autonomous software engineering (successor path: Medium 3.5).
Mistral Small 4
- Context
- 256K / 64K out
- Parameters
- —
- Price
- $0.15 / $0.60
- Speed
- 140 tok/s · 0.2s TTFT
- Modalities
- text, image → text
- License
- Apache 2.0
Hybrid instruct/reasoning/coding Small model under Apache 2.0 - Mistral's high-throughput production default.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.