fullauto.online

← Model leaderboard

Model · stat sheet

Xiaomi MiMo 2.6 Pro

Xiaomi · open weights·open
Overall
86
Rank
#11 / 34
Coding88
Terminal85
Reasoning87
Tool use85
Context85
Speed85
CodingTerminalReasoningTool useContextSpeed

The Pareto-frontier update: V2.5 pricing, a much stronger model. Scored nine days after release — an early read.

What it is

Xiaomi's September 2026 open-weight flagship: the MiMo-V2.6 series update to the V2.5 generation, published with a 1M-token context window and — the point of the release — the V2.5 price list unchanged ($0.43 / $0.87 per 1M tokens on OpenRouter, weights mirrored for local runs). Xiaomi's own release chart (Sept 2026, vendor-reported) puts MiMo-V2.6-Pro at 34.9 on Terminal Bench 4.0 against 1.5 for MiMo-V2.5-Pro; we have not seen an independent re-run of that figure, so treat the direction as real and the magnitude as provisional. Its Artificial Analysis index of 46.3 is the highest number in the 84–86 composite cluster on the board.

How the composite is built

Blended from six axes (Coding 24% · Terminal 20% · Reasoning 20% · Tool use 15% · Context 13% · Speed 8%), it lands at 86 — one point behind the frontier and level with models costing several times more.

Coding — 88

measured on real edit-and-run workloads: multi-file changes, tests passing, diffs that hold up under review.

Terminal — 85

measured on long agentic sessions in a shell: planning, running tools, recovering from failed commands.

Reasoning — 87

measured on planning and hard decompositions — the questions a model has to answer before it acts.

Tool use — 85

measured on calling tools correctly: right arguments, right ordering, no hallucinated functions.

Context — 85

measured on usable context under load: what survives at the top of a long session.

Speed — 85

throughput and latency; higher is faster.

Where it fits

The default open-weights pick for agent work when frontier pricing is unjustified: long coding sessions, harness runs behind OpenCode or similar, batch work where a 1M window matters. It is the first open model on this board where we would not apologise for choosing it over a frontier lab on a hard task.

Limits

Scored 26 Sep 2026, nine days after release — an early read that will move as the model meets more real workloads. The Terminal Bench number above is Xiaomi's own. The value rating leans on an AA index that lands mid-scale; if the index moves, the "Excellent" verdict moves with it.

Price and access

$0.43 in / $0.87 out per 1M tokens ($0.54 blended, 3:1 in:out) — rated Excellent for value: 46.3 index points at $0.54/M blended. Open weights, so local deployment is possible for anyone with the hardware.

Alternatives on this board

Sources

  • OpenRouter: MiMo-V2.6-Pro — pricing as of Sept 2026.
  • Xiaomi MiMo-V2.6 release materials (Sept 2026) — benchmark figures are vendor-reported.
  • Artificial Analysis index, 46.3, as of 25 Sept 2026.

Scores are fullauto.online's composite index (0–100): Coding 24% · Terminal 20% · Reasoning 20% · Tool use 15% · Context 13% · Speed 8%. Editorial, not a vendor benchmark; 2026 tiers are early reads. Last scored 26 Sep 2026 · back to the leaderboard.