LLMs TigerLive LLM benchmarks

MiniMax-M2.5: benchmark scores, price and ranking

MiniMax-M2.5 from MiniMax ranks #484 of 743 LLMs on the LLMs Tiger aggregate benchmark score (46.8), based on 12 benchmark results captured up to 2026-10-13.

Organisation
MiniMax
Type
Open-weight / local model
Open weights
Yes
Released
2026-02-12 (source: curated)
Cheapest API price
$0.270 in / $1.08 out per million tokens (via openrouter, observed 2026-10-13T19:47Z)
Aggregate rank
#484 of 743 (score 46.8)

Relative to other models, MiniMax-M2.5 is strongest on Epoch ECI (#19 of 204) and weakest on CL-bench (#16 of 16).

Across the 11 benchmarks where at least ten models are scored, MiniMax-M2.5 is in the top 10% on 2 and in the top half on 5.

Its cheapest listed input price, $0.270 per million tokens, is lower than that of 62% of the 427 models with a listed price.

MiniMax-M2.5 benchmark results

MiniMax-M2.5 benchmark scores, ranked against all models measured on each benchmark
BenchmarkScoreRankSourceCaptured
Epoch ECI147#19 of 204Epoch AI Benchmarking Hub2026-10-13
SWE-bench Verified75.8%#7 of 73SWE-bench Verified leaderboard2026-10-13
MMLU-Pro80.1%#35 of 136HuggingFace MMLU-Pro leaderboard2026-10-13
GPQA Diamond85.2%#80 of 275HuggingFace GPQA Diamond leaderboard2026-10-13
Terminal-Bench 2.042.7%#11 of 28Epoch AI Benchmarking Hub2026-10-13
SWE-bench Multilingual68.3%#6 of 11SWE-bench Multilingual leaderboard2026-10-13
WebDev Arena1,384#41 of 67Epoch AI Benchmarking Hub2026-10-13
ALE-Bench618#53 of 79Epoch AI Benchmarking Hub2026-10-13
ARC-AGI-24.9%#60 of 87ARC Prize official leaderboard2026-10-13
ProofBench4.0%#42 of 46Epoch AI Benchmarking Hub2026-10-13
CL-bench11.4%#16 of 16Epoch AI Benchmarking Hub2026-10-13
CL-bench Life6.3%#9 of 9Epoch AI Benchmarking Hub2026-10-13

About MiniMax-M2.5

MiniMax-M2.5 is an open-weight model that can run on your own hardware from MiniMax. It was released on 2026-02-12, according to the developer's announcement. The cheapest API listing we track charges $0.270 per million input tokens and $1.08 per million output tokens, through openrouter. We hold 12 benchmark results for it across 3 areas, and 2 are in the top tenth of their boards.

Where MiniMax-M2.5 stands

On our aggregate score, MiniMax-M2.5 is #484 of 743 ranked models, in the lower half. Its score is 46.8; the scale is centred on 50, so it sits 3.2 points below the typical model. That beats 35% of the ranked field. Among the 6 ranked MiniMax models it is #4. The next one up is MiniMax-M2 at 48.7. The next one down is MiniMax-M3 at 45.6.

MiniMax-M2.5, benchmark by benchmark

On Epoch ECI, MiniMax-M2.5 scores 147 and ranks #19 of 204. Only 18 models score higher, led by Qwen 3.8 Max at 156, which is 10 points ahead. It covers Epoch Capabilities Index: Epoch's own aggregate of ~40 benchmarks. The board median is 127, so MiniMax-M2.5 is 19 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

On SWE-bench Verified, MiniMax-M2.5 scores 75.8% and ranks #7 of 73. Only 6 models score higher, led by Claude Opus 4.7 (max) at 83.5%, which is 7.7 points ahead. It covers real GitHub issues resolved. The board median is 56.4%, so MiniMax-M2.5 is 19.4 points above it. Source: SWE-bench Verified leaderboard, captured 2026-10-13.

MiniMax-M2.5 manages 80.1% on MMLU-Pro, good for #35 of 136. Darwin-180B-RSI leads at 88.1%, 8.0 points further up, and 74% of models sit below. The results either side are qwen3-30b-a3b-thinking-2507 (80.9%) and ERNIE-4.5-300B-A47B-PT (78.4%). It covers 57-subject knowledge + reasoning. The board median is 55.3%, so MiniMax-M2.5 is 24.8 points above it. Source: HuggingFace MMLU-Pro leaderboard, captured 2026-10-13.

GPQA Diamond puts MiniMax-M2.5 at 85.2%, #80 of 275. That beats 71% of the models tested, but 10.6 points short of GPT-6 Astra (max) (95.8%). The results either side are GLM-5-FP8 (85.4%) and Darwin-4B-David (85.0%). It covers expert-level science questions. The board median is 71.7%, so MiniMax-M2.5 is 13.5 points above it. Source: HuggingFace GPQA Diamond leaderboard, captured 2026-10-13.

On Terminal-Bench 2.0, MiniMax-M2.5 reaches 42.7%. It ranks #11 of 28, ahead of 61% of the field, and 26.7 points behind Gemini 3 Pro Preview at 69.4%. The results either side are Kimi K2.5 (43.2%) and DeepSeek-V3.2 (Thinking; Novita) (39.6%). It covers real terminal tasks (agent + model). The board median is 37.3%, so MiniMax-M2.5 is 5.4 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

MiniMax-M2.5 lands below the middle on SWE-bench Multilingual: 68.3%, #6 of 11, with 45% of the field scoring higher. Gemini 3 Flash leads at 72.7%; the closest result above is Gemini 3 Pro at 68.7%. It covers real GitHub issues resolved. The median score on this board is also 68.3%. Source: SWE-bench Multilingual leaderboard, captured 2026-10-13.

MiniMax-M2.5 lands below the middle on ALE-Bench: 618, #53 of 79, with 66% of the field scoring higher. GPT-6 Astra (max) leads at 2,951; the closest result above is Qwen3.5 Plus at 622. It covers AtCoder heuristic-optimisation contests (performance rating). The board median is 764, so MiniMax-M2.5 is 146 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

MiniMax-M2.5 lands below the middle on ARC-AGI-2: 4.9%, #60 of 87, with 68% of the field scoring higher. GPT-6 Astra (max) leads at 95.0%; the closest result above is GLM-5 at 4.9%. It covers abstract reasoning, ARC Prize public eval set. The board median is 22.8%, so MiniMax-M2.5 is 17.9 points below it. Source: ARC Prize official leaderboard, captured 2026-10-13.

ProofBench is a weak spot. MiniMax-M2.5 scores 4.0%, #42 of 46, behind 89% of the models measured. The leader, Claude Fable 5.1 (max), reaches 100.0%. It covers writing mathematical proofs. The board median is 21.0%, so MiniMax-M2.5 is 17.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

CL-bench is a weak spot. MiniMax-M2.5 scores 11.4%, #16 of 16, behind 94% of the models measured. The leader, GPT-5.4 (xhigh), reaches 27.9%. It covers learning new rules from context. The board median is 17.7%, so MiniMax-M2.5 is 6.3 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

The full table above also lists the other 2 results, including boards with fewer than ten models, where a rank says little.

Strengths and weaknesses

Software engineering: solid, ahead of 56% of the field on average across 4 benchmarks. Its best board here is SWE-bench Verified (#7 of 73) and its weakest is ALE-Bench (#53 of 79). Reasoning, maths and science: uneven, ahead of 46% of the field on average across 4 benchmarks. Its best board here is MMLU-Pro (#35 of 136) and its weakest is ProofBench (#42 of 46). General knowledge and chat: uneven, ahead of 45% of the field on average across 3 benchmarks. Its best board here is Epoch ECI (#19 of 204) and its weakest is CL-bench (#16 of 16).

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so MiniMax-M2.5 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 deepseek-chat ($0.280) and Gemini 3.5 Flash-Lite (High) ($0.300).

Release history

The MiniMax model released before it was MiniMax-M2.1 on 2025-12-23, which scores 49.9. The next one was MiniMax-M2.7 on 2026-03-18, at 45.6.

Reading these numbers

The 12 scores on this page come from 6 sources: Epoch AI Benchmarking Hub (7), ARC Prize official leaderboard (1), HuggingFace GPQA Diamond leaderboard (1), HuggingFace MMLU-Pro leaderboard (1). 2 of them come from an official leaderboard and 10 from datasets or third-party runs.

Models ranked nearby

More from MiniMax

Compare it on the live leaderboard → · How the aggregate score works