MiMo-V2.6-Pro
Xiaomi's flagship open-weight model (MIT license), a trillion-parameter model that reads text, images, video and audio with a 1M-token context window. It has the highest Artificial Analysis Intelligence Index of any open-weight model in this list and the lowest cost per task among models scoring 40 or more.
XiaomiLanguage modelReleased September 2026Data checked October 18, 2026
This model is not in the current people's rankings: it is either new and still collecting votes, or ranked below the top 12. See the AI leaderboard for the models people rate highest.
On paper
Fixed-test rankings from the benchmarks tab: Value #1, Reasoning #8, Speed #19.
Strengths and weaknesses
- Top open-weight score on the Artificial Analysis Intelligence Index (46.3)
- Best value in this list: $0.13 per Artificial Analysis task
- Very low price ($0.435 / $0.87 per million tokens)
- Slow output on Artificial Analysis (48 tokens per second)
- Fewer LMArena votes than older models, so its arena scores have wide confidence intervals
Best for
- The best value open model
- Agents on a budget
- Long multimodal documents
Every score we have
Each bar compares this model with the best result any model on this page has on the same test.
- LMArena text arena (human votes)1,479.2Arena score 1479.2 (95% CI 1469.8 to 1488.6); 4,075 votes; entry "mimo-v2.6-pro"; position 27 of 416 on the full board
- LMArena creative writing1,444.7Arena score 1444.7 (95% CI 1423.6 to 1465.8); 793 votes; entry "mimo-v2.6-pro"; position 52 of 414 on the full board
- LMArena WebDev (coding)1,624.8Arena score 1624.8 (95% CI 1614.9 to 1634.7); 4,763 votes; entry "MiMo-V2.6-Pro"; position 20 of 144 on the full board
- LMArena vision1,246Arena score 1246.0 (95% CI 1231.4 to 1260.7); 1,778 votes; entry "mimo-v2.6-pro"; position 59 of 165 on the full board
- LMArena Agent Arena0.040IPS score 0.0395 (95% CI 0.0287 to 0.0504); 27,455 sessions; entry "MiMo-V2.6-Pro"; position 13 of 52 on the full board
- Artificial Analysis Intelligence Index46.346.3 on the Artificial Analysis Intelligence Index v4.3.2 (AA shows 46), setting "MiMo-V2.6-Pro"
- Humanity's Last Exam (Artificial Analysis)49.4%49.4% on Humanity's Last Exam (AA, 2,158 text-only questions), setting "MiMo-V2.6-Pro"
- Terminal-Bench 4.0 (Artificial Analysis run)34.8%34.8% on Terminal-Bench 4.0 as run by AA (mini-swe-agent harness), setting "MiMo-V2.6-Pro"
- Long-context reasoning (AA-LCR)86.3%86.3% on AA-LCR (long-context reasoning), setting "MiMo-V2.6-Pro"
- AA-Omniscience, accuracy minus hallucination8.48.4 on the AA-Omniscience Index (knowledge accuracy minus hallucination, range -100 to 100), setting "MiMo-V2.6-Pro"
- Output speed, tokens per second48 t/s48 output tokens per second (AA median), row "MiMo-V2.6-Pro"
- Cost per Artificial Analysis test task$0.13$0.13 to run one AA Intelligence Index task, row "MiMo-V2.6-Pro"
Compare with
- Claude Opus 5.5Anthropic's flagship for long-running agentic coding and knowledge work, and the model Anthropic suggests starting with for most jobs.
- Claude Sonnet 5.5Anthropic's mid-priced model, which Anthropic describes as its best mix of speed and intelligence.
- Claude Fable 5.1Anthropic's top tier for the hardest reasoning and long-horizon agent work, which Anthropic recommends when Opus 5.5 still falls short.
- Claude Haiku 5.5Anthropic's smallest and fastest current model, built for high-volume, latency-sensitive jobs like classification, extraction and routing.