Back to models
Intelligence Index
Standard setting
9.3
Measured · Artificial Analysis v4.3.2- API price from
- $0.50 in / $1.50 out / 1M tokOpenRouter · Mistral
- Context window
- 256K tok
Performance profile
ReasoningArchived source
68.0%
GPQA DiamondCodingArchived source
36.6%
SciCodeCost efficiencyArchived source
88.8pts / $
Intelligence ÷ eval costCapabilities
Context window
Modalities
Measured setting
Integration
Reasoning trade-offs
Source (opens in a new tab)Mistral Large 3
Cost per task (USD) · log scale
Mistral Large 3Standard
- Intelligence Index
- 9.3
- Cost / task
- $0.1044
- Evaluation time
- 1.2m
Pricing & access
USD per 1M tokens
- Input
- $0.50
- Output
- $1.50
- Cached input
- $0.05
Run locally
GGUF
Context
Estimated memory
~417.5 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.
512 GiB RAMwithout GPU offload
Memory budgetOne text chat
- Weights
- 379 GiB
- Context cache
- 0.5 GiB16-bit cache
- App allowance
- 37.9 GiB
- Download
- 407 GB9 files · keep together+5.1 GB for images (opens in a new tab)
- Disk space
- ~459 GBIncludes image file
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-Large-3-675B-Instruct-2512-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-Large-3-675B-Instruct-2512-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)4.2%36.6%0.0%36.0%406 Elo
Humanity’s Last Exam
Frontier reasoning- 1Claude Opus 5.5Max61.4%
- 2Claude Fable 5.1Max59.1%
- 3Gemini 4 ArgonHigh57.1%
- 4Claude Sonnet 5.5Max55.0%
- 5GPT-6 AstraMax54.7%
- 6Muse Wonder FullMax53.6%
- 7GPT-6.1 SolMax52.9%
- 8Muse WonderMax50.1%
- 9MiMo-V2.6-ProReasoning49.4%
- 10Muse Spark 1.3Max48.7%
Peers use their best measured result on this benchmark.