English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

24 Hours, 9.4x: How MiniMax M3 Turned Itself Into an Engineer

Forum topic · 小凯 · 2026-06-01

Summary

MiniMax M3, released in Shanghai on June 1, 2026, is the first Chinese open-source model claimed to combine three frontier capabilities: agent-grade coding, 1M-token context, and native multimodal training. Its MiniMax Sparse Attention (MSA) architecture inverts conventional sparse attention by gathering queries at the KV-block level, delivering 9x prefilling and 15x decoding speedups at 1M context with roughly 1/20 the per-token compute of its predecessor. The technical report highlights three autonomous engineering tasks: reproducing an ICLR 2025 paper in ~12 hours with 18 commits and 23 figures; writing an FP8 GEMM kernel from scratch in 24 hours, raising hardware utilization from 7.6% to 71.3% (9.4x speedup) across 147 benchmark submissions; and autonomously training base models over 12 hours, scoring 0.37 versus GPT-5.5's 0.39 and Opus 4.7's 0.42. M3 scores 59.0% on SWE-Bench Pro, and ships with the MiniMax Code agent product and aggressive subscription pricing. Open-source weights were promised within 10 days.

> Published on zhichai.net — English translation of a Chinese forum post. The post is dated June 1, 2026, Shanghai.

MiniMax M3 is not framed as "another Chinese GPT clone." The claim: it is the first domestic (Chinese) model to simultaneously reach usable levels on three capabilities that overseas closed-source frontier models all have — Coding, 1M context, and native multimodality — while choosing to open-source. More strikingly, it publicly documented throwing itself into a 24-hour CUDA optimization ordeal, writing an FP8 GEMM kernel from scratch and pushing hardware utilization from 7.6% to 71.3%.

Key points

1. Three "tickets" to the frontier

  • Coding: agent-level programming (understanding requirements, using terminals, tool calls, multi-turn iteration), measured by SWE-Bench Pro, Terminal Bench, MCP Atlas.
  • Long context: 1M tokens (~1.5M words), required for sustained agent workflows.
  • Native multimodality: trained on text, image, and video from step 0, not bolted-on vision.
  • The post argues previous Chinese models each had at most two of the three; M3 is the first to hold all three.

    2. MSA: MiniMax Sparse Attention

  • Attention is O(n²); at 1M tokens full attention costs 64x that of 128K.
  • Existing approaches: DSA (fixed-interval key sampling), MoBA (block selection), FlashAttention (IO-efficient but still quadratic).
  • MSA inverts the flow: KV blocks as the outer loop, gathering the queries that hit them ("outer gather Q"). KV blocks are read contiguously once and serve multiple queries — cache-friendly and high compute-to-memory ratio.
  • MiniMax reports MSA is 4x+ faster than Flash-Sparse-Attention and flash-moba under M3's head configuration, with capability parity to full attention in ablations.
  • At 1M context: 1/20 per-token compute, 9x prefilling speedup, 15x decoding speedup (end-to-end measured).
  • 3. Native multimodality

  • Interleaved data (text-image alternating in-sequence) proved more impactful than paired data — cross-modal understanding emerges from sequence-level interleaving, not additive fusion.
  • Data pipeline rebuilt to target ~100 trillion training tokens.
  • Three stress tests

    2.1 Paper reproduction (12 hours, 18 commits, 23 figures)

    Given an ICLR 2025 Outstanding Paper ("Learning Dynamics of LLM Finetuning"), M3 autonomously reproduced it: ~12 hours runtime, 18 git commits, 23 experimental figures, matching SFT-stage predicted-probability trends, observing the DPO "squeezing" effect, and validating the Extend mitigation.

    2.2 CUDA kernel optimization (24 hours, 147 submissions, 9.4x speedup)

  • Starting materials: a task description, a benchmark script, a non-runnable Triton skeleton, no reference implementation.
  • 24 hours: 147 benchmark submissions, 1959 tool calls, fully autonomous.
  • Six optimization stages: baseline (7.6% utilization) → autotune config generation → bottleneck diagnosis → CUDA Graph integration → persistent kernel rewrite → host-side scheduling.
  • Result: peak utilization 7.6% → 71.3%, 9.4x speedup.
  • Notably: competing models typically gave up within ~30 submissions; M3's best solution came at submission #145, persisting through multiple plateaus — attributed to long-context attention allocation.
  • 2.3 Self-directed model training (12 hours)

    Given four pre-trained-only Base models, M3 autonomously handled data synthesis, training, evaluation, and iteration (AIME2025, BFCL, GPQA Main, GSM8K, HumanEval). Final score 0.37, below Opus 4.7 (0.42) and GPT-5.5 (0.39), clearly ahead of others.

    Benchmark table

    | Benchmark | M3 | Comparison | |------|--------|--------| | SWE-Bench Pro | 59.0% | above GPT-5.5, Gemini 3.1 Pro; near Opus 4.7 | | Terminal Bench 2.1 | 66.0% | — | | SWE-fficiency | 34.8% | — | | KernelBench Hard | 28.8% | — | | MCP Atlas | 74.2% | — | | SVG-Bench | — | above Opus 4.7 | | OmniDocBench | — | above Gemini 3.1 Pro | | Claw-Eval | — | highest score | | PostTrainBench | 0.37 | below Opus 4.7 (0.42), GPT-5.5 (0.39) |

    Product, pricing, open-source

    MiniMax Code (agent product, built on OpenCode and Pi Agent): Agent Team workflows, a Producer + Verifier adversarial harness, multi-day autonomous runs, and Computer Use triggered by phone voice commands. Compared to Claude Code's JS-based Dynamic Workflows, MiniMax Code emphasizes deep reflection and continuous correction.

    Pricing: Plus ¥49/mo (600M tokens), Max ¥119/mo (1.8B), Ultra ¥469/mo (5.5B) — roughly 15x Claude's token volume at comparable prices. Caveat: token count ≠ model quality; thinking-mode latency and API stability aren't visible in pricing.

    Open-source: MiniMax promised to release the technical report and open model weights within 10 days.

    Open questions raised by the post

  • How close is "close to Opus 4.7" on SWE-Bench Pro, exactly?
  • The 7.6% baseline in the CUDA task may be a deliberately weak starting point — "from broken to usable" differs from "from good to better."
  • Is flash-moba a fair comparison baseline? Full MSA technical details are pending.
  • Will the 10-day open-source promise be honored?

Conclusion

The post argues M3's real significance is a complete capability stack for "engineer-type AI": understanding existing knowledge (paper reproduction), solving known problems (kernel optimization), and exploring open problems (training models). It trails Opus 4.7 and GPT-5.5 on the hardest tier, but the gap is narrowing — and if the open weights ship on time, M3 could become key infrastructure for developers studying agentic engineering AI.

References cited in the post

1. MiniMax Research (2026). MiniMax M3: Frontier Coding, 1M Context, Native Multimodal. minimaxi.com/blog. 2. Vaswani et al. (2017). Attention Is All You Need. NeurIPS. 3. Gu & Dao (2024). Mamba: Linear-Time Sequence Modeling with Selective State Spaces. ICML. 4. Dao et al. (2022). FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness. NeurIPS. 5. Garg et al. (2022). What Can Transformers Learn In-Context? A Case Study of Simple Function Classes. NeurIPS.

*Note: this is a translation of a community post; claims reflect the original author's reporting of MiniMax's announcements, not independent verification by zhichai.net.*

Tags

#minimax#m3#sparse-attention#long-context#coding-agent#cuda#open-source#benchmark

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177980709