What are the verified coding and reasoning benchmarks for Minimax M3?

Minimax M3 achieves SWE-bench Verified: 58% (Standard), LiveCodeBench: 50.5%. Verified sources: Artificial Analysis Intelligence Index v4.3.2 (Artificial Analysis), MiniMax-M3 Model Card (MiniMax).

What is the recommended Swarm multi-agent topology for Minimax M3?

Swarm recommends Minimax M3 as Coder Subagent (Tier 2: Documentation & Test Specialist). Generates comprehensive documentation and test scaffolding quickly. Best paired with: Claude Opus 5.5 or Qwen 3.8 Max.

How does Minimax M3 compare to rival frontier coding models on SWE-bench and pricing?

Minimax M3 achieves SWE-bench Verified: 58% at $0.25/$1 per 1M tokens. Direct peer comparison: K2 Horizon 375B A23B (59.8% SWE-bench, $0.4/$1.6); Qwen 3.8 27B (62% SWE-bench, $0.2/$0.8); NVIDIA Nemotron 3 Ultra (58.2% SWE-bench, $0.5/$1.5); Inkling (54.5% SWE-bench, $0.15/$0.6).

Does Swarm isolate Minimax M3 code edits from other agents?

Yes. Swarm provisions isolated Git worktrees for every session and subagent, preventing file race conditions and keeping your main working tree clean until verification passes.

Can Minimax M3 be assigned to specific subagent roles?

Yes. In ~/.swarm/config.json, you can configure Minimax M3 as Primary/Planner (DAG task decomposition), Coder (implementation in isolated worktrees), or Finder (read-only symbol search).

Fireworks AI model

Minimax M3

Q: How do I run Minimax M3 with Swarm?

Configure Minimax M3 via CLI using: swarm ctl model set --provider fireworks --model fireworks/minimax-m3, or switch models interactively inside the terminal TUI with /models.

MiniMax-M3 is a native multimodal model with 1M context running ~428B parameters and ~23B activated parameters. It brings native multimodality. enabling deeper semantic fusion across text, image, and video. M3 also introduces MiniMax Sparse Attention (MSA) to improve long context efficiency, achieving frontier-level performance across long-horizon agentic benchmarks, excelling in both coding and cowork.

Provider: Fireworks AI
Availability: Available
Model ID: minimax-m3
Snapshot: 2026-10-07

Minimax M3 Benchmarks & Agentic Accuracy

Verified primary research & developer preference metrics

Primary Sources Verified

SWE-bench Verified

58%

StandardMiniMax AI ↗

LiveCodeBench

50.5%

ThinkingLiveCodeBench ↗

LMSYS Coding Elo

1320Elo

Coding ArenaLMSYS Arena ↗

Pricing ($ / 1M tokens)

$0.25/ $1

$0.05 cached readStandard API Pricing

What This Means in Real Software Development

Achieves 29 on the Artificial Analysis Intelligence Index v4.3.2.
Scores 30 on the Finance & Accounting Index.
Delivers 85 TPS throughput at an efficient $0.25 cost per task.
Specialized in conversational coding, documentation generation, and test synthesis.

Swarm Multi-Agent Topology

Tier 2: Documentation & Test SpecialistRole: Coder Subagent

Best Paired With: Claude Opus 5.5 or Qwen 3.8 Max

Generates comprehensive documentation and test scaffolding quickly.

Git Worktree Isolation Advantage:

Works in isolated docs/tests branches without impacting core logic.

What Developers & Practitioners Are Saying

Real-world field observations, production strengths, and developer consensus

Field Consensus

"A dependable and cost-effective subagent choice for documentation writing, test scaffolding, and simple code refactors."

Verified Real-World Strengths

Inexpensive $0.25 cost per task with 85 TPS throughput
Clean documentation and natural language explanations
Reliable API availability and low latency

Production Caveats & Gaps

Struggles on very complex compiler or kernel-level codebases
Best managed under a strong Lead planner

Primary Sources & Methodology Citations

Artificial Analysis Intelligence Index v4.3.2 — Artificial Analysis (September 2026)
MiniMax-M3 Model Card — MiniMax (June 2026)

Head-to-Head Frontier Comparisons: Minimax M3 vs Rivals

Standardized, side-by-side benchmark & economics comparison against peer models

Flagship Comparative Matrix

Model	Role / Tier	SWE-bench	LiveCodeBench	LMSYS Elo	AIME	Input / Output
Minimax M3Fireworks AICurrent	Tier 2: Documentation & Test Specialist	58%	50.5%	1320	—	$0.25 / $1
K2 Horizon 375B A23BIndependent Compare Minimax M3 vs K2 Horizon 375B A23B →	Tier 2: MoE Systems Specialist	59.8%	52%	1330	—	$0.4 / $1.6
Qwen 3.8 27BQwen Compare Minimax M3 vs Qwen 3.8 27B →	Tier 2: Open-Weights Specialist	62%	54.5%	1342	—	$0.2 / $0.8
NVIDIA Nemotron 3 UltraNVIDIA Compare Minimax M3 vs NVIDIA Nemotron 3 Ultra →	Tier 2: Accelerated Systems Specialist	58.2%	53%	1365	77%	$0.5 / $1.5
InklingIndependent Compare Minimax M3 vs Inkling →	Tier 2: Code Search & Discovery Specialist	54.5%	47%	1305	—	$0.15 / $0.6

Explore All AI Model Comparisons Hub (Minimax M3 & Flagship Peers) →

Run Minimax M3 in Swarm

Verified CLI workflow & multi-agent architecture

Quickstart Guide →

bash — swarm run

$swarm ctl model set --provider fireworks --model fireworks/minimax-m3

# Or switch interactively inside the terminal TUI:

>/models

Subagent Role Matrix

Assign Minimax M3 to specialized system roles in ~/.swarm/config.json:

Primary / Planner: Orchestrates full task DAGs and multi-stage pipelines.
Coder Subagent: Generates code and unit tests inside private Git worktrees.
Finder Subagent: Performs read-only code search, symbol discovery, and mapping.

System Agent Roles →Why Swarm Architecture →

← All Fireworks AI models

Public catalog snapshot. Catalog availability is not a real-time provider status or an access guarantee.