Mistral Large 3
- Input
- $0.50/M tok
- Output
- $1.50/M tok
- Blended
- $0.75/M tok · 3:1
- Speed
- 77tok/s
- Context
- 256Ktokens
- Cost per task
- ~$0.63/taskest.
Previous-generation open-weights flagship. Main release is FP8 (card: "FP8 on a single node of B200s or H200s"; BF16 provided "if needed").
Rankings
Coding
Verified#114 / 137rank by quality25.8±6.5quality / 1003/7 benchmarks measuredWriting & chat
Verified#89 / 162rank by quality38.6±3.3quality / 1004/5 benchmarks measuredVision
Verified#80 / 98rank by quality26.8±4.8quality / 1003/6 benchmarks measuredHard reasoning
Verified#159 / 218rank by quality24.6±3.3quality / 1004/4 benchmarks measuredAgents
Verified#147 / 187rank by quality15.7±4.3quality / 1004/7 benchmarks measured
Self-hostingEstimated
How we estimate →What it would cost to run these open weights yourself on rented GPUs, compared with the API price above.
- Hardware
- 8× B200 180GB
- FP8 weights
- Throughput
- ~20,662 tok/s
- many requests batched
- Price
- $0.28–2.49 /M tok
- busy → light use
- On your own machine
- Mac Studio M5 Ultra 512GB (int4, ~41 tok/s single-stream)
- single consumer GPU or Mac
Assumptions (5)
- FP8 weights (675B params, 41B active per token) + 25% KV-cache headroom ≈ 844 GB
- 8× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
- ~20,662 output tok/s aggregate at batch 591 (bandwidth-bound); MoE compute scales with active params
- Blended 3:1 input:output; prefill ~207,439 tok/s
- Low = 75% utilization at the low GPU price; high = 20% at the high price
Benchmark scores
~ italic, dashed = no published score yet, estimated from related benchmarks.
Coding
SWE-bench VerifiedEstimated
~#17 of 21 (estimated)~67.1%
- #1Claude Opus 596.0%
- #2Claude Sonnet 585.2%
- #16Nemotron 3 Ultra70.7%
- ~#17Mistral Large 3Estimated67.1%
- #17Intern-S2-Preview (35B-A3B)64.0%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#23 of 24 (estimated)~45.0%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #22Muse Glimmer51.2%
- ~#23Mistral Large 3Estimated45.0%
- #23Nex-N2.5-Mini43.8%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#42 of 46 (estimated)~38.6%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #41Laguna S 2.140.4%
- ~#42Mistral Large 3Estimated38.6%
- #42Nex-N2.5-Mini36.1%
No published score — estimated from related benchmarks
Terminal-Bench 4.0Verified
#63 of 1010.0%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #63Ministral 3 8B0.0%
- #63Mistral Large 30.0%
- #63Mistral Medium 3.50.0%
artificialanalysis.ai · 2026-09-24
LiveCodeBenchVerified
#20 of 2746.5%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #19Falcon-H1-34B-Instruct49.7%
- #20Mistral Large 346.5%
- #21Devstral 244.8%
artificialanalysis.ai · 2026-09-24
SciCodeVerified
#92 of 11836.6%
- #1Claude Fable 5.267.7%
- #2Claude Opus 5.566.9%
- #91G9v3-39A5B36.8%
- #92Mistral Large 336.6%
- #92Qwen3.6 35B A3B36.6%
artificialanalysis.ai · 2026-09-24
AA Coding Agent IndexEstimated
~#25 of 26 (estimated)~27.1
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #24GPT-6 Luna31.1
- ~#25Mistral Large 3Estimated27.1
- #25Muse Spark Open26.3
No published score — estimated from related benchmarks
Writing & chat
IFBenchVerified
#97 of 13936.2%
- #1MiniMax-M382.9%
- #2Qwen3.8 2.4T A95B82.8%
- #96Command A36.5%
- #97Mistral Large 336.2%
- #98Nanbeige4.1-3B35.4%
artificialanalysis.ai · 2026-09-24
LMArena Text, Non-English (Elo)Verified
#39 of 561397 Elo
- #1Gemini 4 ArgonFew votes1512
- #2Claude Opus 5.51497
- #38Nemotron 3 Ultra1405
- #39Mistral Large 31397
- #39Qwen3.5 122B A10B1397
lmarena.ai · 2026-09-24
LMArena Text (Elo)Verified
#41 of 561413 Elo
- #1Gemini 4 Argon1525
- #2Claude Opus 5.51507
- #40Claude Haiku 4.51415
- #41Mistral Large 31413
- #42GPT-5.4 nano1402
lmarena.ai · 2026-09-24
#22 of 251412 Elo
- #1GPT-6 Astra2173
- #2Claude Fable 5.12162
- #21Gemini 3.1 Pro1491
- #22Mistral Large 31412
- #23Gemma 4 31B1368
eqbench.com · 2026-09-24
Vision
MMMU-ProVerified
#66 of 8755.7%
- #1Claude Opus 5.587.7%
- #2GPT-6 Astra86.9%
- #65Mistral Small 456.8%
- #66Mistral Large 355.7%
- #67Magistral Small 1.255.5%
artificialanalysis.ai · 2026-09-24
CharXiv ReasoningEstimated
~last of 12 (estimated)below measured range (<74.5%)
- #1Gemini 3.8 Flash86.2%
- #2Kimi K384.8%
- #11Gemini 3.5 Flash-Lite74.5%
- ~#12Mistral Large 3Estimated<74.5%
No published score — estimated from related benchmarks
OmniDocBench v1.5Estimated
~last of 8 (estimated)below measured range (<75.8%)
No published score — estimated from related benchmarks
GDP.pdf (AA)Verified
#41 of 471.2%
- #1Muse Wonder Full32.6%
- #2GPT-6 Astra31.0%
- #41Ministral 3 14B1.2%
- #41Mistral Large 31.2%
- #43Command A+1.0%
artificialanalysis.ai · 2026-09-24
Hard reasoning
GPQA DiamondVerified
#119 of 19268.0%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #118K2-V268.1%
- #119Mistral Large 368.0%
- #120Claude Haiku 4.567.2%
artificialanalysis.ai · 2026-09-24
Humanity's Last ExamVerified
#183 of 1984.2%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #183LFM2 24B A2B4.2%
- #183Mistral Large 34.2%
- #185Command A4.0%
artificialanalysis.ai · 2026-09-24
Agents
τ²-bench (Telecom)Verified
#73 of 11924.6%
- #1Step 3.7 Flash98.5%
- #2DeepSeek-V4-Pro-0813-nvfp4-DSpark98.3%
- #71Ministral 3 3B24.9%
- #73Mistral Large 324.6%
- #74Devstral Small 223.4%
artificialanalysis.ai · 2026-09-24
OSWorld-VerifiedEstimated
~#12 of 15 (estimated)~57.6%
- #1Kimi K384.8%
- #2Qwen3.8 27B84.3%
- #11Muse Glimmer65.9%
- ~#12Mistral Large 3Estimated57.6%
- #12Nemotron 3.5 Super VL NVFP452.3%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~last of 5 (estimated)below measured range (<47.5%)
No published score — estimated from related benchmarks
BrowseCompEstimated
~#19 of 22 (estimated)~50.4%
- #1Atria Dawn Preview92.5%
- #2GPT-6 Astra91.5%
- #18Nemotron 3.5 Bolt 74B-A9B70.0%
- ~#19Mistral Large 3Estimated50.4%
- #19Mistral Medium 3.548.6%
No published score — estimated from related benchmarks