MiniMax-M3
open weights★ featuredTop 5 cheapest providers (Adjusted $/task)
Within the active global provider, residency and confidentiality filters.
| # | Provider | Platform | Raw input $/1M | Raw output $/1M | Adjusted $/task |
|---|---|---|---|---|---|
| 1 | CoreWeave | OpenRouter | $0.230 | $0.960 | |
| 2 | GMICloud | OpenRouter | $0.240 | $0.960 | |
| 3 | DeepInfra | OpenRouter | $0.280 | $1.10 | |
| 4 | Nebius | Nebius | $0.300 | $1.20 | |
| 5 | Venice | OpenRouter | $0.300 | $1.20 |
Benchmarks
ⓘ Intelligence.ai (formerly DesignArena) publishes this result at product/family scope without an effort setting. It is attached exactly once to minimax-m3::default, the deterministic family representative used by collapsed comparisons; this does not assert that Intelligence.ai tested this specific effort setting.
MODEL EVIDENCE
Benchmark sheet
22 / 73 registered benchmark versions covered · 26 observations including existing index snapshots.
Composite retains Coding Agent v1.4 · source 2026-09-09
The five Composite inputs remain unchanged. Its Coding Agent input is the median across complete harness results in the retained v1.4 snapshot from 2026-09-09. Artificial Analysis now publishes v1.5, with different components. Explore current v1.5 separately. Dated snapshot labels on other indices identify unversioned source captures, not a verified semantic version.
Profile signals
1 threshold-crossing signal flagged · 24 eligible benchmark families
How flags are calculated
Heuristic screen, not statistical significance: benchmark families are correlated and source uncertainty is unknown. Peer evidence needs ≥ 20 independently measured matched configurations from ≥ 10 distinct model families. A flag needs a directed population z-score of magnitude ≥ 1.5 and a gap of ≥ 1.5 from the leave-one-benchmark-family-out mean z in the same direction, over ≥ 5 other benchmark families.
- unusually strong IFBench (AA single-turn) snapshot-2026-09-10
Why
Observed: 0.82857 fractionPeer mean 0.48006 · peer sd 0.17139 · n 450 · families 331Directed z 2.033 · baseline z 0.449 · gap 1.584 · profile n 230.82857 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗Evidence
Axis: IFBench (AA single-turn) · snapshot-2026-09-10 · Published boardExact value:0.828571428571429fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recordedObservation id:aa:277f939a-985b-4b37-859d-b3eabc7c0b26:ifbench
Protocol-compatible measured/vendor divergences
No verified protocol-compatible vendor/measured pair is available for this model; agreement cannot be assessed.
Agentic
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| AA-AnalystAgent Version snapshot-2026-09-10 | Published board Tests analyst tasks using agentic Python execution across fourteen domains. | 0.1 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: AA-AnalystAgent · snapshot-2026-09-10 · Published board Exact value: 0.1 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:analystAgent |
| AA-Briefcase Version snapshot-2026-09-10 | Published board Tests multi-week professional knowledge-work projects with linked tasks and large source collections. | 1096.03 Elomeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: AA-Briefcase · snapshot-2026-09-10 · Published board Exact value: 1096.03 EloObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:briefcaseBreakdown.overall.elo |
| GDPval-AA v2 Version 2 | Published board Tests professional knowledge-work deliverables across occupations using AA's Stirrup harness. | 1303.96 Elomeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: GDPval-AA v2 · 2 · Published board Exact value: 1303.96 EloObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:gdpval |
| Harvey LAB-AA Version snapshot-2026-09-10 | Published board Tests legal-work deliverables across practice areas on Harvey's private task set. | 0.88406 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: Harvey LAB-AA · snapshot-2026-09-10 · Published board Exact value: 0.884064264003474 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:harveyLab |
| Terminal-Bench Hard (AA) Version 74221fb | Published board Tests a pinned 44-task hard subset of terminal-based work using Terminus 2. | 0.42424 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: Terminal-Bench Hard (AA) · 74221fb · Published board Exact value: 0.424242424242424 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:terminalbenchHard |
| Terminal-Bench v2.1 (AA) Version 2.1 | Published board Tests terminal-based work on the 89-task verified refresh using Terminus 2. | 0.65169 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: Terminal-Bench v2.1 (AA) · 2.1 · Published board Exact value: 0.651685393258427 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:terminalbenchV21 |
| Terminal-Bench v4.0 (AA) Version 4.0 | Published board Tests terminal-based work on the 66-task release using mini-SWE-agent v2.4.6. | 0.0202 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: Terminal-Bench v4.0 (AA) · 4.0 · Published board Exact value: 0.0202020202020202 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:terminalbenchV40 |
Coding
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| SciCode (AA subproblems) Version 1.0.1 | Published board Tests scientific Python programming with scientist-annotated background information. | 0.47106 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: SciCode (AA subproblems) · 1.0.1 · Published board Exact value: 0.471064814814815 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:scicode |
| AA Coding Index Version snapshot-2026-09-10 (unversioned) | Published board Published AA index from the retained API snapshot. No verified semantic version was supplied. | 58.6 pointsmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: AA Coding Index · snapshot-2026-09-10 (unversioned) · Published board Exact value: 58.6 pointsObserved: 2026-09-10 · publication date: not recorded Observation id: legacy:aa_coding_index:minimax-m3::defaultProtocol: inspect · file data/raw/artificialanalysis.json |
| DesignArena Frontend Version snapshot-2026-09-10 (unversioned) | Published board Published Elo. Exact source attachment is retained; fewer than 200 battles excludes a row from radar normalization and anomaly peers. | 1221 Elomeasured2597 battlesobserved 2026-09-10designarena.ai ↗ EvidenceAxis: DesignArena Frontend · snapshot-2026-09-10 (unversioned) · Published board Exact value: 1221 EloObserved: 2026-09-10 · publication date: not recorded Observation id: legacy:frontend:minimax-m3::defaultSource: https://www.designarena.ai/ Protocol: inspect · file data/raw/designarena.json |
| DesignArena Full-Stack Version snapshot-2026-09-10 (unversioned) | Published board Published Elo. Exact source attachment is retained; fewer than 200 battles excludes a row from radar normalization and anomaly peers. | 1208 Elomeasured2275 battlesobserved 2026-09-10designarena.ai ↗ EvidenceAxis: DesignArena Full-Stack · snapshot-2026-09-10 (unversioned) · Published board Exact value: 1208 EloObserved: 2026-09-10 · publication date: not recorded Observation id: legacy:fullstack:minimax-m3::defaultSource: https://www.designarena.ai/ Protocol: inspect · file data/raw/designarena.json |
Instruction-following
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| IFBench (AA single-turn) Version snapshot-2026-09-10 | Published board Tests precise single-turn instructions with deterministic rule checks. | 0.82857 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: IFBench (AA single-turn) · snapshot-2026-09-10 · Published board Exact value: 0.828571428571429 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:ifbench |
Knowledge
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| Humanity's Last Exam (AA text-only) Version snapshot-2026-09-10 | Published board Tests expert-level knowledge on AA's text-only Humanity's Last Exam subset. | 0.38971 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: Humanity's Last Exam (AA text-only) · snapshot-2026-09-10 · Published board Exact value: 0.389712696941613 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:hle |
| AA-Omniscience Index Version snapshot-2026-09-10 | Published board Tests factual reliability while rewarding correct answers and penalizing hallucinations. | 1.35 pointsmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: AA-Omniscience Index · snapshot-2026-09-10 · Published board Exact value: 1.35 pointsObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:omniscience |
Long-context
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| GDP.pdf (AA) Version snapshot-2026-09-10 | Published board Tests professional reasoning over long PDFs with AA document preparation and grading. | 0.098 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: GDP.pdf (AA) · snapshot-2026-09-10 · Published board Exact value: 0.098 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:gdpPdfAllPass |
| AA-LCR v1.1 Version 1.1 | Published board Tests reasoning across multiple long documents with corrected answer keys and grading. | 0.83 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: AA-LCR v1.1 · 1.1 · Published board Exact value: 0.83 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:lcr |
| MLCR-AA Version snapshot-2026-09-10 | Published board Tests medical-record synthesis and reasoning across long, fragmented case documents. | 0.17222 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: MLCR-AA · snapshot-2026-09-10 · Published board Exact value: 0.172222222222222 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:mlcrOverall |
Reasoning
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| AA Intelligence Index Version snapshot-2026-09-10 (unversioned) | Published board Published AA index from the retained API snapshot. No verified semantic version was supplied. | 29.6 pointsmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: AA Intelligence Index · snapshot-2026-09-10 (unversioned) · Published board Exact value: 29.6 pointsObserved: 2026-09-10 · publication date: not recorded Observation id: legacy:aa_intelligence_index:minimax-m3::defaultProtocol: inspect · file data/raw/artificialanalysis.json |
Science
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| CritPt (AA) Version snapshot-2026-09-10 | Published board Tests research-level physics reasoning with Python, symbolic and numerical answers. | 0.03714 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: CritPt (AA) · snapshot-2026-09-10 · Published board Exact value: 0.0371428571428571 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:critpt |
| GPQA Diamond (AA) Version snapshot-2026-09-10 | Published board Tests graduate-level biology, physics and chemistry knowledge on the Diamond subset. | 0.92929 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: GPQA Diamond (AA) · snapshot-2026-09-10 · Published board Exact value: 0.929292929292929 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:gpqa |
Tool-use
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| AutomationBench-AA Version 1.0.6 | Published board Tests multi-app SaaS workflows through REST tools on a held-out AutomationBench split. | 0.21251 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: AutomationBench-AA · 1.0.6 · Published board Exact value: 0.21251257600749407 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:automationBenchPartialScore |
| EnterpriseOps-Gym-AA Version snapshot-2026-09-10 | Published board Tests enterprise workflows through MCP tools against resettable application environments. | 0.3211 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: EnterpriseOps-Gym-AA · snapshot-2026-09-10 · Published board Exact value: 0.3210981796478663 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:enterpriseOpsGym |
| τ²-Bench Telecom (AA) Version snapshot-2026-09-10 | Published board Tests dual-control telecom agents that coordinate tool use with a simulated user. | 0.88889 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: τ²-Bench Telecom (AA) · snapshot-2026-09-10 · Published board Exact value: 0.888888888888889 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:tau2 |
| τ³-Banking (AA) Version 1.0.1 | Published board Tests banking support agents that retrieve policies and change account state through tools. | 0.15258 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: τ³-Banking (AA) · 1.0.1 · Published board Exact value: 0.152577319587629 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:tauBanking |
Vision
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| MMMU Pro (AA) Version snapshot-2026-09-10 | Published board Tests multimodal understanding using challenging ten-option questions. | 0.78555 fractionmeasuredobserved 2026-09-10artificialanalysis.ai ↗ EvidenceAxis: MMMU Pro (AA) · snapshot-2026-09-10 · Published board Exact value: 0.785549132947977 fractionObserved: 2026-09-10T21:47:16.627Z · publication date: not recorded Observation id: aa:277f939a-985b-4b37-859d-b3eabc7c0b26:mmmuPro |
Writing
| Benchmark / version | Evaluation group | Score / provenance |
|---|---|---|
| LLM Creative Story-Writing Benchmark Version snapshot-2026-09-10 | Published board Pairwise comparison of short stories written to the same constrained creative briefs, with LLM evaluator choices combined into a relative comparison score. | 0.1 pointsmeasuredobserved 2026-09-10raw.githubusercontent.com ↗ EvidenceAxis: LLM Creative Story-Writing Benchmark · snapshot-2026-09-10 · Published board Exact value: 0.1 pointsObserved: 2026-09-10T21:53:00Z · publication date: not recorded Observation id: public:bda1d85b21ce37948959f4cf |
Missing coverage · 51 benchmark versions
No result does not mean a zero, or that the model was never tested. Collection failures and disputed versions retain their distinct status.
- AIME 2025 (AA) · 2025
Unknown: no published result matched to this configuration.
- APEX-Agents-AA · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- Artificial Analysis Coding Agent Index v1.4 · 1.4
Unknown: no published result matched to this configuration.
- Artificial Analysis Coding Agent Index v1.5 · 1.5
Unknown: no published result matched to this configuration.
- ITBench-AA · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- LiveCodeBench (AA historical field) · snapshot-2026-09-10
contested: Exact evaluation task window is unknown.
- Aider Polyglot · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- ARC-AGI-1 · 1
Unknown: no published result matched to this configuration.
- ARC-AGI-2 · 2
Unknown: no published result matched to this configuration.
- ARC-AGI-3 · 3
Unknown: no published result matched to this configuration.
- BuzzBench · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- CritPt · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- EQ-Bench 4 · 4
Unknown: no published result matched to this configuration.
- EQ-Bench Creative Writing v3 · 3
Unknown: no published result matched to this configuration.
- Judgemark v4 · 4
Unknown: no published result matched to this configuration.
- EQ-Bench Longform Creative Writing · v1.11
Unknown: no published result matched to this configuration.
- Slop Score · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- HELMET · snapshot-2026-09-10
manual required: Suite uses heterogeneous task/context metrics; no universal headline is published.
- Humanity's Last Exam (HLE) · snapshot-2026-09-10
manual required: README sample output is not attributable to a model and is not a current result.
- IFBench · snapshot-2026-09-10
manual required: Captured README specifies evaluation, not a named-model result.
- Japanese-RP-Bench (Aratako) · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- Japanese-RP-Bench v2 (tegnike fork) · 2
Unknown: no published result matched to this configuration.
- LiveBench · 2026-06-25
Unknown: no published result matched to this configuration.
- LongBench v2 · 2
Unknown: no published result matched to this configuration.
- LLM Divergent Thinking Creativity Benchmark · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- Elimination Game Benchmark · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- MMMU · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- OTIS Mock AIME 2024-2025 · 2024-2025
contested: Captured July-2025 result predates the registered exact-match grading protocol; withheld.
- PingPong English (unadjusted mean) · 2
Unknown: no published result matched to this configuration.
- RP-Bench Community Elo · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- RULER · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- SciCode · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- SimpleBench · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- The Slop Index · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- SlopBench · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- Spiral-Bench · 1.2
Unknown: no published result matched to this configuration.
- SWE-bench Multilingual · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- SWE-bench Multimodal · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- SWE-Bench Pro Public · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- SWE-bench Verified · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- τ²-bench · 2
Unknown: no published result matched to this configuration.
- τ³-Banking · 1.0.1
Unknown: no published result matched to this configuration.
- τ³-Voice · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- Terminal-Bench · 4.0
Unknown: no published result matched to this configuration.
- Towards AI Editorial Writing · snapshot-2026-09-10
manual required: Private protocol and visually reported X values require xplainervideo capture and primary evidence review.
- UGI NatInt · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- UGI Willingness · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- UGI Writing · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- UGI · snapshot-2026-09-10
Unknown: no published result matched to this configuration.
- Vending-Bench 2 · 2
Unknown: no published result matched to this configuration.
- WeirdML · 2
Unknown: no published result matched to this configuration.
Token offers by platform — Adjusted $/task
Adjusted costs are modeled USD per task: AA output tokens × OpenRouter usage I/O (Chutes global fallback). These are general usage and benchmark proxies for coding-agent work. Missing AA data assumes 1,000 output tokens/task; unknown cache hit assumes 0%; unmeasured additional cache writes assume 0 tokens. Click any underlined price for exact inputs, dates and assumptions. Raw list prices use the selected fixed input/output blend, in USD per million tokens.
13 offers within the active global filters; “—” means the catalog is active but no public token price is available.
OpenRouter (10)
| CoreWeave | globalcoreweave/fp4 | $0.230 raw in $/1M | $0.960 raw out $/1M | |
| GMICloud | globalgmicloud/fp8 | $0.240 raw in $/1M | $0.960 raw out $/1M | |
| DeepInfra | globaldeepinfra/fp8 | $0.280 raw in $/1M | $1.10 raw out $/1M | |
| Venice | globalvenice/fp8 | $0.300 raw in $/1M | $1.20 raw out $/1M | |
| Together | globaltogether | $0.300 raw in $/1M | $1.20 raw out $/1M | |
| Parasail | globalparasail/fp8 | $0.300 raw in $/1M | $1.20 raw out $/1M | |
| AtlasCloud | globalatlas-cloud/fp8 | $0.300 raw in $/1M | $1.20 raw out $/1M | |
| Novita | globalnovita/fp8 | $0.300 raw in $/1M | $1.20 raw out $/1M | |
| SambaNova | globalsambanova | $0.600 raw in $/1M | $2.40 raw out $/1M | |
| ModelRun | globalmodelrun/fp4 | $0.750 raw in $/1M | $3.00 raw out $/1M |
Nebius (1)
| Nebius | us | $0.300 raw in $/1M | $1.20 raw out $/1M |
Azure AI Foundry (1)
| Azure AI Foundry | usfireworks | $0.330 raw in $/1M | $1.32 raw out $/1M |
TensorX (1)
| TensorX | euEU | $0.400 raw in $/1M | $2.00 raw out $/1M |