Ctrl K
Back to models

Mistral

Mistral Medium 3.5

  • Released
  • Open weights
  • Modified MIT License
Compare
Intelligence Index

High setting

14.2#61 of 83measured models
Measured · Artificial Analysis v4.3.2
API price from
$1.50 in / $7.50 out / 1M tokOpenRouter · Mistral
Context window
256K tok

Performance profile

ReasoningRank #67 of 107
74.8%
GPQA Diamond
CodingRank #63 of 94
40.2%
SciCode
Cost efficiencyRank #53 of 67
28.3pts / $
Intelligence ÷ eval cost

Capabilities

Context window

Modalities

Measured setting

Integration

Reasoning trade-offs

Source (opens in a new tab)
Intelligence Index v4.3.2
Mistral Medium 3.5
Cost per task (USD) · log scale
Mistral Medium 3.5High
Intelligence Index
14.2
Cost / task
$0.502
Evaluation time
2.5m

Pricing & access

API route
OpenRouter · MistralAPI
Tier & contextStandard

USD per 1M tokens

Input
$1.50
Output
$7.50

Run locally

GGUF
Context
Estimated memory

~79.5 GiB

RAM + VRAM

Estimate assumptions

One text chat at the selected context, with a 16-bit cache and app allowance. Operating system and image processing are extra; optimized caches can use less.

96 GiB RAMwithout GPU offload

Memory budgetOne text chat
Weights
69.8 GiB
Context cache
2.8 GiB16-bit cache
App allowance
7 GiB
Download
74.9 GB3 files · keep together+5.4 GB for images (opens in a new tab)
Disk space
~94 GBIncludes image file
Available software
Ollama (opens in a new tab)LM Studio (opens in a new tab)Jan (opens in a new tab)KoboldCpp (opens in a new tab)llama.cpp (opens in a new tab)
Get GGUF (opens in a new tab)
Set up in Ollama · Q4_K_M · 8K context

Save the following as Modelfile. The context matches the estimate above.

FROM hf.co/unsloth/Mistral-Medium-3.5-128B-GGUF:Q4_K_M
PARAMETER num_ctx 8192

Then run these commands in your terminal:

ollama create llmcurrent-local -f Modelfile
ollama run llmcurrent-local

Ollama import instructions (opens in a new tab) · For LM Studio or Jan, open the app’s model search, enter unsloth/Mistral-Medium-3.5-128B-GGUF, choose Q4_K_M, and set context to 8,192 tokens in the app’s model settings. Software links above open each app’s setup guide.

Images need extra memoryvLLM setup (opens in a new tab)

Benchmarks

Full leaderboardSource (opens in a new tab)

Humanity’s Last Exam

Frontier reasoning

Mistral Medium 3.5 · High#80 of 126

  1. 1Claude Opus 5.5Max61.4%
  2. 2Claude Fable 5.1Max59.1%
  3. 3Gemini 4 ArgonHigh57.1%
  5. 77Qwen3.5 9BReasoning14.9%
  6. 78Qwen3.5 Omni PlusStandard14.9%
  7. 79K2 Horizon 3.7BReasoning13.9%
  8. 80Mistral Medium 3.5High13.8%
  9. 81EXAONE 4.5 33BReasoning12.9%
  10. 82Qwen3 Next 80B A3B (Reasoning)Reasoning12.6%
  11. 83Command A+Reasoning12.0%

Peers use their best measured result on this benchmark.

Alternatives

Each model at its highest-scoring measured setting · Intelligence Index v4.3.2 and its evaluation cost per task.