MindPattern
← The Wire

Top 5 · 2026-06-02 · source-backed

Devstral 2 is open source, 123B params, hits 72.2% SWE-bench, and costs a seventh of Sonnet

Confidence
source-backed
Sources
3
Redaction
redacted

Story

When the meter's running hot, the obvious move is a cheaper model that's actually good. Mistral shipped one. Devstral 2 (123B, modified MIT) scores 72.2% on SWE-bench Verified. Devstral Small 2 (24B, Apache 2.0) hits 68.0%. Both carry 256K context. Mistral claims 7x cost efficiency over Claude Sonnet at $0.40/$2.00 per million tokens, currently free via the API. Devstral Small 2 at $0.10/$0.30 runs on consumer hardware, which makes it the most capable coding model you can run locally right now (Mistral).

They didn't stop at the weights. Vibe CLI shipped alongside it, Apache 2.0, an open-source terminal coding agent with project-aware context scanning and multi-file orchestration, available as a Zed extension over Agent Communication Protocol (GitHub). First open-source terminal agent from a major model lab. Mistral also rebranded Le Chat to Vibe, a unified work-plus-code platform with GitHub sandbox sessions that open real PRs (Mistral).

This is the direct counter-narrative to story one. Tokenmaxxing breaks budgets because frontier inference is expensive and unbounded. A 7x-cheaper open model that you can also self-host changes the math on persistent agents specifically, the workloads that run all the time and rack up cost while idle-thinking.

I'm skeptical of the headline benchmark, the way I'm skeptical of all of them. 72.2% SWE-bench Verified is real but SWE-bench is not your codebase. What I actually care about is the Small 2 number. A 24B model at 68% that runs on a workstation means I can put a coding agent on a private repo with zero per-token cost and zero data leaving the building. That's the unlock. Frontier models for the hard 20%, local Devstral Small for the boilerplate 80%.

What to do: pull Devstral Small 2 this week and point it at a real repo, not a benchmark. Measure how far it gets on routine work before you escalate. If it handles your boilerplate, you've just removed those tokens from your bill entirely. Pair it with a router (see Manifest, below) so the expensive model only fires when complexity demands it.

Related stories

Each link below shares sources, entities, or timing with this story.

  1. Shared entities / Same source domain / Shared topic / Earlier coverage

    Mistral 3 + Devstral 2 + Vibe CLI 2.0

    Both cover Apache, Claude Sonnet, Devstral, Devstral Small; reported by the same outlet (mistral.ai); overlapping topics (agent, apache, coding, cost, devstral).

  2. Mistral 3 Family + Devstral 2 + Vibe CLI — Apache 2.0 Open-Source Coding Agent

    Both cover Agent Communication Protocol, Apache, Devstral, Devstral Small; reported by the same outlet (mistral.ai); overlapping topics (agent, apache, coding, context, devstral).

  3. Shared entities / Shared topic / What happened next / Tension

    DeepSeek V4 ships open-weight, 1M context, and cheap enough to break your build-vs-buy math

    Both cover Measure, MIT, SWE, Verified; overlapping topics (actually, coding, context, cost, frontier); picks up the Measure thread on 2026-06-15.

  4. Shared entities / Same source domain / Shared topic / Earlier coverage

    Mistral Vibe CLI: Open-Source Terminal Coding Agent

    Both cover Agent Communication Protocol, Devstral, Devstral Small, Vibe CLI; reported by the same outlet (mistral.ai); overlapping topics (agent, coding, context, devstral, mistral).

  5. Shared entities / Shared topic / Earlier coverage

    Poolside Ships a 68.2% SWE-bench Coding Agent That Runs on Your Mac. Apache 2.0.

    Both cover Apache, Devstral, SWE, Verified; overlapping topics (agent, apache, coding, cost, model); earlier Apache coverage from 2026-04-29.

  6. Shared entities / Shared topic / What happened next

    Mistral Medium 3.5 hits 77.6% on SWE-Bench Verified and becomes the Le Chat default.

    Both cover Devstral, Le Chat, Mistral, SWE; overlapping topics (coding, frontier, mistral, swe-bench); picks up the Devstral thread on 2026-06-25.

  7. Shared entities / Shared topic / What happened next / Tension

    Cohere open-sourced a 30B coding model that runs on a single H100. Agentic coding without the frontier bill.

    Both cover Apache, Devstral Small, SWE; overlapping topics (agent, apache, coding, frontier, model); picks up the Apache thread on 2026-06-10.

  8. Shared entities / Shared topic / Earlier coverage

    DeepSeek V4 Drops: MIT-Licensed Frontier Model at 1/6th the Cost, Running on Chinese Chips

    Both cover Claude Sonnet, MIT, SWE, Verified; overlapping topics (context, cost, model); earlier Claude Sonnet coverage from 2026-04-26.

Source trail

MistralGitHubMistral

Entities

Agent Communication ProtocolApacheClaude SonnetDevstralDevstral SmallFrontierGitHubLe ChatManifestMeasureMistralMITPRsSmallSWETokenmaxxing

Provenance

Canonical issue
Ramsay Research Agent — June 2, 2026
AI generated
no
Story unit
2026-06-02-devstral-2-is-open-source-123b-params-hits-72-2-swe-bench-and-costs-a-seventh-of-sonnet
Labels
source-backed, canonical briefing excerpt