Skip to content
Model Pareto

Mistral Medium 3.5

Mistral AIOpen weightsmistral.ai ↗
Input
$1.50/M tok
Output
$7.50/M tok
Blended
$3.00/M tok · 3:1
Speed
149tok/s
Context
256Ktokens
Cost per task
~$3.2/taskest.

Open weights (modified MIT).

Rankings

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
4× H100 80GB
FP8 weights
Throughput
~2,947 tok/s
many requests batched
Price
$0.54–5.20 /M tok
busy → light use
On your own machine
Mac Studio M5 Max 128GB (int4, ~7 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (128B params) + 25% KV-cache headroom ≈ 160 GB
  • 4× H100 80GB at $2.69–$6.88/GPU-hour on-demand (2026-09-24)
  • ~2,947 output tok/s aggregate at batch 86 (bandwidth-bound)
  • Blended 3:1 input:output; prefill ~14,611 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

~#23 of 24 (estimated)~51.1%
  1. #1Claude Fable 5.293.0%
  2. #2Claude Opus 5.589.9%
  3. #22Muse Glimmer51.2%
  4. ~#23Mistral Medium 3.5Estimated51.1%
  5. #23Nex-N2.5-Mini43.8%
No published score — estimated from related benchmarks
~#25 of 26 (estimated)~30.4
  1. #1Claude Opus 5.561.7
  2. #2Claude Opus 559.7
  3. #24GPT-6 Luna31.1
  4. ~#25Mistral Medium 3.5Estimated30.4
  5. #25Muse Spark Open26.3
No published score — estimated from related benchmarks

Writing & chat

Vision

Hard reasoning

Agents