MiniMax-M3: benchmark scores, price and ranking
MiniMax-M3 from MiniMax ranks #566 of 743 LLMs on the LLMs Tiger aggregate benchmark score (45.6), based on 28 benchmark results captured up to 2026-10-13.
- Organisation
- MiniMax
- Type
- Open-weight / local model
- Open weights
- Yes
- Released
- 2026-06-01 (source: epoch)
- Cheapest API price
- $0.230 in / $0.960 out per million tokens (via wandb, observed 2026-10-13T19:47Z)
- Aggregate rank
- #566 of 743 (score 45.6)
Relative to other models, MiniMax-M3 is strongest on Epoch ECI (#17 of 204) and weakest on LiveBench Mathematics (#17 of 17).
Across the 25 benchmarks where at least ten models are scored, MiniMax-M3 is in the top 10% on 1 and in the top half on 12.
Its cheapest listed input price, $0.230 per million tokens, is lower than that of 66% of the 427 models with a listed price.
MiniMax-M3 benchmark results
About MiniMax-M3
MiniMax-M3 is an open-weight model that can run on your own hardware from MiniMax. It was released on 2026-06-01, according to Epoch AI. The cheapest API listing we track charges $0.230 per million input tokens and $0.960 per million output tokens, through wandb. We hold 28 benchmark results for it across 4 areas, and 1 is in the top tenth of its board.
Where MiniMax-M3 stands
On our aggregate score, MiniMax-M3 is #566 of 743 ranked models, in the lower half. Its score is 45.6; the scale is centred on 50, so it sits 4.4 points below the typical model. That beats 24% of the ranked field. Among the 6 ranked MiniMax models it is #5. The next one up is MiniMax-M2.5 at 46.8. The next one down is MiniMax-M2.7 at 45.6.
MiniMax-M3, benchmark by benchmark
On Epoch ECI, MiniMax-M3 scores 147 and ranks #17 of 204. Only 16 models score higher, led by Qwen 3.8 Max at 156, which is 9 points ahead. It covers Epoch Capabilities Index: Epoch's own aggregate of ~40 benchmarks. The board median is 127, so MiniMax-M3 is 20 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On GPQA Diamond, MiniMax-M3 reaches 90.9%. It ranks #28 of 275, ahead of 90% of the field, and 4.9 points behind GPT-6 Astra (max) at 95.8%. The results either side are GLM-5.3 (max) (90.9%) and Qwen3.7 Max (90.9%). It covers expert-level science questions. The board median is 71.7%, so MiniMax-M3 is 19.2 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
MiniMax-M3 manages 55.0% on Surface Evolver, good for #4 of 13. Kimi K3 (max) leads at 93.0%, 38.0 points further up, and 69% of models sit below. The results either side are Grok 4.5 (high) (74.4%) and GLM-5.3-Flash (max) (52.5%). It covers geometry optimisation with a solver. The board median is 46.2%, so MiniMax-M3 is 8.8 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
WebDev Arena puts MiniMax-M3 at 1,487, #24 of 67. That beats 64% of the models tested, but 334 points short of Claude Opus 5.5 (max) (1,820). The results either side are GLM-5.1 (1,508) and Qwen 3.6 Max (Preview) (1,479). It covers building web apps, crowd-voted (Elo). The board median is 1,433, so MiniMax-M3 is 53 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On DTBench, MiniMax-M3 reaches 78.9%. It ranks #57 of 146, ahead of 61% of the field, and 20.0 points behind Claude Opus 5.5 (max) at 98.9%. The results either side are qwen3.5-35b-a3b (80.0%) and Qwen3-235B-A22B-Instruct-2507 (78.4%). It covers dynamic tool-use benchmark. The board median is 68.8%, so MiniMax-M3 is 10.1 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
MiniMax-M3 manages 71.1% on OTIS Mock AIME, good for #77 of 189. Claude Fable 5.1 (max) leads at 100.0%, 28.9 points further up, and 59% of models sit below. The results either side are claude-sonnet-4-20250514_32K (71.1%) and gemini-2.5-flash-preview-05-20 (70.8%). It covers competition maths (AIME 2024-25 style). The board median is 57.8%, so MiniMax-M3 is 13.3 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
LiveBench IF is a weak spot. MiniMax-M3 scores 57.5%, #15 of 17, behind 82% of the models measured. The leader, gemini-3.1-pro-preview-high, reaches 79.1%. It covers LiveBench release 2026_06_25. The board median is 65.2%, so MiniMax-M3 is 7.7 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
LiveBench Reasoning is a weak spot. MiniMax-M3 scores 74.5%, #15 of 17, behind 82% of the models measured. The leader, claude-opus-4-8-xhigh-effort, reaches 89.7%. It covers LiveBench release 2026_06_25. The board median is 81.1%, so MiniMax-M3 is 6.6 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
LiveBench Coding is a weak spot. MiniMax-M3 scores 68.2%, #16 of 17, behind 88% of the models measured. The leader, GPT-5.2 Codex, reaches 83.6%. It covers LiveBench release 2026_06_25. The board median is 76.5%, so MiniMax-M3 is 8.3 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
LiveBench Mathematics is a weak spot. MiniMax-M3 scores 76.9%, #17 of 17, behind 94% of the models measured. The leader, GPT-5.5 (xhigh), reaches 95.9%. It covers LiveBench release 2026_06_25. The board median is 88.2%, so MiniMax-M3 is 11.3 points below it. Source: LiveBench 2026_06_25 official CSV, captured 2026-10-13.
The full table above also lists the other 18 results, including boards with fewer than ten models, where a rank says little.
Strengths and weaknesses
Software engineering: uneven, ahead of 34% of the field on average across 6 benchmarks. Its best board here is WebDev Arena (#24 of 67) and its weakest is LiveBench Coding (#16 of 17). Reasoning, maths and science: uneven, ahead of 45% of the field on average across 7 benchmarks. Its best board here is GPQA Diamond (#28 of 275) and its weakest is LiveBench Mathematics (#17 of 17). Agents and tool use: uneven, ahead of 43% of the field on average across 2 benchmarks. Its best board here is DTBench (#57 of 146) and its weakest is APEX-Agents (#12 of 16). General knowledge and chat: uneven, ahead of 48% of the field on average across 10 benchmarks. Its best board here is Epoch ECI (#17 of 204) and its weakest is LiveBench IF (#15 of 17).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so MiniMax-M3 costs less than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on GPT-5 mini (high) ($0.250) and MiniMax-M2.7 ($0.250).
Release history
The MiniMax model released before it was MiniMax-M2.7 on 2026-03-18, which scores 45.6.
Reading these numbers
The 28 scores on this page come from 4 sources: Epoch AI Benchmarking Hub (19), LiveBench 2026_06_25 official CSV (7), LMArena (1), SWE-rebench leaderboard (1). 7 of them come from an official leaderboard and 21 from datasets or third-party runs. The leaderboard also lists 1 other reasoning-effort setting of this model, which redirect here. This page shows the highest.
MiniMax-M3 is also listed at 1 other reasoning-effort setting; this page shows the highest-effort row, which is the one the leaderboard keeps by default.
Models ranked nearby
- #563 qwen-plus-2025-01-25 (45.7)
- #564 MiMo-V2-Flash (45.6)
- #565 Llama 4 Scout Instruct (45.6)
- #567 open-mistral-nemo-2407 (45.6)
- #568 MiniMax-M2.7 (45.6)
- #569 ollama/mistral-nemo:12b-instruct-2407-q4_K_M (45.6)
More from MiniMax
- MiniMax-M1-80k (#160, 50.9)
- MiniMax-M2.1 (#228, 49.9)
- MiniMax-M2 (#330, 48.7)
- MiniMax-M2.5 (#484, 46.8)
- MiniMax-M2.7 (#568, 45.6)
Compare it on the live leaderboard → · How the aggregate score works