Mistral Medium 3.5

Mid-tier

Mistral Medium 3.5 entered public preview on May 3, 2026 and is the most practical entry in the open-weight coding tier. It is a dense 128 billion parameter model with a 256K context window, released under a modified MIT license, priced at $1.50 per million input tokens and $7.50 per million output. It posts 77.6 percent on SWE-Bench Verified, which is frontier-adjacent coding performance at a size that fits on a single node. The reason it matters more than its benchmark rank suggests is deployability: quantized to 4-bit it needs roughly 72GB, so it runs on one H100. Compare that to Kimi K3 at about 1,450GB and eight B200-class cards for the same license freedom. Strong enough to matter, small enough to actually run, and licensed permissively enough to ship.

Input Price

$1.50

per 1M tokens

Output Price

$7.50

per 1M tokens

Context Window

256K

tokens

Released

2026-05

Open source

Capabilities

textcodetool-usereasoning

Key Strengths

✓77.6 percent on SWE-Bench Verified at 128B dense
✓Runs on a single H100 at 4-bit (roughly 72GB)
✓256K context window
✓Modified MIT license permitting commercial self-hosting
✓Best capability-per-GPU ratio in the open-weight coding tier

Best For

▸Single-node self-hosted coding agents
▸European data-residency deployments
▸Cost-controlled IDE and refactor automation
▸Teams that need open weights they can actually run

Benchmark Scores

Benchmark	Score	Description
SWE-bench	77.6	Real-world software engineering tasks from GitHub issues (SWE-bench Verified)

Scores sourced from public benchmark datasets. See full benchmark leaderboard for all models.

Pricing Details

Input tokens

$1.50

per 1M tokens

Output tokens

$7.50

per 1M tokens

Estimated cost per 1K requests

$5.25

~1K input + ~500 output tokens avg

Prices are subject to change. Check the official documentation for current pricing. See the cost calculator for detailed estimates.

Open Source Model

Mistral Medium 3.5 is free to download and self-host under the Modified MIT. Hosted API pricing varies by provider (e.g., Together, Fireworks, Groq). See our open source LLM guide for deployment options.