vs
2:0
overall capability wins (0 ties across 2 scored). Devstral 2 leads overall.
- Pricing3:0
- Speed2:0
- Specs2:0
Verdict
Devstral 2 offers ~2.0x larger context (256K vs 128K); Devstral 2 allows ~1.6x more max output (64K vs 40K); Devstral 2 ships open weights (Apache 2.0).
- Devstral 2 offers ~2.0x larger context (256K vs 128K)
- Devstral 2 allows ~1.6x more max output (64K vs 40K)
- Devstral 2 ships open weights (Apache 2.0)
- Devstral 2 is ~3.4x cheaper on a blended token basis than Magistral Medium
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Devstral 2
- $0.90
- Magistral Medium
- $3.25
Devstral 2 is about 3.6x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Mistral
Mistral
Family
Mistral
Mistral
License
Apache 2.0
Proprietary
Open weights
Yes
No
Release
Dec 1, 2025
Jul 1, 2025
Knowledge cutoff
-
-
API / provider
Mistral
Mistral
Modalities
text → text
text, image → text
Specs
Context window
256K
128K
Max output
64K
40K
Parameters
123B
—
Pricing
Input $/1M
$0.40
$2
Output $/1M
$2
$5
Blended $/1M (3∶1)
$0.80
$2.75
Speed
tok/s
85
60
TTFT (s)
0.35
0.8
Reasoning
MMLU-Pro
76.2
—
GPQA Diamond
59.4
—
Humanity's Last Exam
3.6
—
AIME 2025
36.7
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
20.8
—
AA-LCR v1.1
32.3
—
CritPt
—
—
MMMU-Pro
—
—
IFBench
38.1
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
44.8
—
Terminal-Bench 2.1
30.3
—
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
32.8
—
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 | Magistral Medium | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Mistral | Mistral | — |
| Family | Mistral | Mistral | — |
| License | Apache 2.0 | Proprietary | — |
| Open weights | Yes | No | — |
| Release | Dec 1, 2025 | Jul 1, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Mistral | Mistral | — |
| Modalities | text → text | text, image → text | — |
| Specs | |||
| Context window | 256K | 128K | A +128K |
| Max output | 64K | 40K | A +24K |
| Parameters | 123B | — | — |
| Pricing | |||
| Input $/1M | $0.40 | $2 | A +$1.60 |
| Output $/1M | $2 | $5 | A +$3 |
| Blended $/1M (3∶1) | $0.80 | $2.75 | A +$1.95 |
| Speed | |||
| tok/s | 85 | 60 | A +25 |
| TTFT (s) | 0.35 | 0.8 | A +0.45 |
| Reasoning | |||
| MMLU-Pro | 76.2 | — | — |
| GPQA Diamond | 59.4 | — | — |
| Humanity's Last Exam | 3.6 | — | — |
| AIME 2025 | 36.7 | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 20.8 | — | — |
| AA-LCR v1.1 | 32.3 | — | — |
| CritPt | — | — | — |
| MMMU-Pro | — | — | — |
| IFBench | 38.1 | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 44.8 | — | — |
| Terminal-Bench 2.1 | 30.3 | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 32.8 | — | — |
| 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 | — | — | — |
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).
Magistral Medium
- Context
- 128K / 40K out
- Parameters
- —
- Price
- $2 / $5
- Speed
- 60 tok/s · 0.8s TTFT
- Modalities
- text, image → text
- License
- Proprietary
Mistral's thinking/reasoning Premier model for transparent multilingual domain reasoning.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.
These two models have no overlapping published benchmarks in our dataset. Compare specs and pricing instead.