Alex Getmanisyarat awal AI
MS

MiniMax M3

MiniMax

Dikeluarkan Dikemas kini
Tetingkap konteks
1M
1,000,000 token
Input / Output
imej, teks, video → teks

Garis masa

  1. · Versi pemberat · pautan bukti
Harga input / output · USD / 1 juta token
ModelPautan buktiHarga input / output
minimax-m3
standard · teks
opencode-go$0.30 / $1.20
  1. · Pautan bukti · GitHub search · MiniMax · commits · rodrigueztaylor87/rod · @rodrigueztaylor87
  2. · Pautan bukti · GitHub search · MiniMax · commits · SilverKnightKMA/paseo-plugins · @SilverKnightKMA
  3. · Pautan bukti · GitHub search · MiniMax · commits · SilverKnightKMA/paseo-plugins · @SilverKnightKMA
    Petikan
    now carry the FULL pi reference '<pi-provider>/<model>' (magpie/minimax/MiniMax-M3),
  4. · Pautan bukti · GitHub search · MiniMax · commits · Lore-Hex/quill-router · @jperla
  5. · Pautan bukti · GitHub search · MiniMax · pulls · xiaoY233/Chat2API · @octo-patch
    Petikan
    feat(minimax): add MiniMax-M3 model entry
  6. · Pautan bukti · GitHub search · MiniMax · pulls · vllm-project/vllm · @Vincent-PRO-AI
    Petikan
    [Model][MiniMax-M3] SM120 Blackwell NVFP4 KV cache Triton reader
  7. · Pautan bukti · GitHub search · MiniMax · pulls · vllm-project/vllm · @Pb314314
    Petikan
    use sparse MLA (dense MLA, DeepSeek-V4 / V4.1, MiniMax-M3, non-MLA models) are not affected; the reason for each is in "Per-model check".
  8. · Pautan bukti · GitHub search · MiniMax · pulls · vllm-project/vllm · @cjackal
    Petikan
    - Pixtral, Cohere Compass and MiniMax-M3 also call `get_cu_seqlens_padding_logics`, so every encoder-CG model whose `cu_seqlens` comes from `maybe_recompute_cu_seqlens` takes its padding from one place. Without
  9. · Pautan bukti · GitHub search · MiniMax · pulls · vllm-project/vllm · @MrGeva
    Petikan
    [Perf][MiniMax-M3] Two-shot fused MoE finalize + all-reduce + norm + MXFP8 up to 512 tokens
  10. · Pautan bukti · GitHub search · MiniMax · pulls · vllm-project/vllm · @MrGeva
    Petikan
    [Perf][MiniMax-M3] Raise the fused MoE finalize + all-reduce cap from 32 to 128 tokens
  11. · Pautan bukti · GitHub search · MiniMax · pulls · vllm-project/vllm · @afriedri
    Petikan
    short prompts at high concurrency is higher on MiniMax-M3 (TP8: up to +21% at 1024 tokens, conc 64; TP4: +14% at 1024
  12. · Pautan bukti · GitHub search · MiniMax · pulls · vllm-project/vllm · @wolfcomos
    Petikan
    Add `minimax_m3_msa_decode_fp16_dot`, an opt-in optimization for MiniMax-M3 Triton sparse decode on SM100-family GPUs. It uses FP16 dot operands with FP32 accumulation and selects splits targeting
  13. · Pautan bukti · GitHub search · MiniMax · pulls · viztor/dsh-opencode-patch · @viztor
    Petikan
    | `opencode-go/minimax-m3` | info | `[KNOWN_DEFECT]` toggle-only model offered Minimal…Max | Off / On (D1) |
  14. · Pautan bukti · GitHub search · MiniMax · pulls · thequanhcm/quanpt1 · @thequanhcm
  15. · Pautan bukti · GitHub search · MiniMax · pulls · tenstorrent/tt-metal · @mdragulaTT
  16. · Pautan bukti · GitHub search · MiniMax · pulls · tenstorrent/tt-metal · @mdragulaTT
    Petikan
    The MiniMax-M3 vision encoder on one Blackhole chip. It turns one image, as the media module's `prepare` resizes it, into its vision tokens: bf16 `[n_tokens,
  17. · Pautan bukti · GitHub search · MiniMax · pulls · sgl-project/sglang · @alphabetc1
  18. · Pautan bukti · GitHub search · MiniMax · pulls · openJiuwen-ai/jiuwenswarm · @octo-patch
  19. · Pautan bukti · GitHub search · MiniMax · pulls · jundot/omlx · @begcdn
    Petikan
    On M2 Ultra (`applegpu_g14d`) the MiniMax-M3 sparse prefill path crashes once a prompt crosses 4096 tokens: `msa_prefill_supported` gates only on shape (`D=128`, `block_size=128`, `qhead_per_kv=16`), so
  20. · Pautan bukti · GitHub search · MiniMax · pulls · iv-org/invidious · @Fijxu
  21. · Pautan bukti · GitHub search · MiniMax · pulls · inference-sim/inference-sim · @sriumcp
    Petikan
    the substrate's layer 8 and 9 price changes (MiniMax-M3 routers replicated in fp32, MXFP8 scale bytes, dsv4-nvfp4 MTP experts at MXFP4, a shared lm_head for
  22. · Pautan bukti · GitHub search · MiniMax · pulls · inference-sim/blis-vllm · @sriumcp
    Petikan
    (`vllm/model_executor/models/config.py:43-55`) for every graph that is not DeepSeek-V4; MiniMax-M3's pool at the same memory is 72645 blocks, not 72893, because the catalog graph now holds
  23. · Pautan bukti · GitHub search · MiniMax · pulls · inference-sim/blis-vllm · @sriumcp
    Petikan
    the figures on the H200 rows and the MiniMax-M3 points are development-split, in-sample numbers.
  24. · Pautan bukti · GitHub search · MiniMax · pulls · goonercbot/hermes-agent · @goonercbot
    Petikan
    - Real configured MiniMax-M3 request through corrected Hermes adapter returned exactly `RECOVERY_OK`, end_turn, original request unchanged.
  25. · Pautan bukti · GitHub search · MiniMax · pulls · ghuntley/how-to-build-a-coding-agent · @octo-patch
    Petikan
    Anthropic-compatible defaults to the standalone workshop agents, with MiniMax-M3.1-Flash as the default and MiniMax-M3 as the alternative.
  26. · Pautan bukti · GitHub search · MiniMax · pulls · boelabs/ai-gateway · @jhisdev
    Petikan
    - `bun run test` fails: `vercel.test.ts` pins `vercel/minimax/minimax-m3` reasoning as `fixed ["high"]`; the regenerated snapshot now says `openai_effort ["none","high"]`. Tests may not be edited by
  27. · Pautan bukti · GitHub search · MiniMax · pulls · ashhart/TensorFold · @ashhart
    Petikan
    CUDA decode attention (`window_attention`, shared by Qwen3.5/3.8, Llama, MiniMax-M3, Spark 2.5, Step5 and LFM): `window_rows` merged every chunk's partial for every row, 128 chunks at
  28. · Pautan bukti · GitHub search · MiniMax · pulls · anomalyco/models.dev · @rjuez00
    Petikan
    - `MiniMax-M3` is no longer listed and a request using that ID returns `model=MiniMax-M3.1-Flash`, so the old entry is removed rather than kept as an
  29. · Pautan bukti · GitHub search · MiniMax · pulls · ROCm/aiter · @zhiding512
    Petikan
    On gfx1250, every MiniMax-M3 sparse attention call (decode, MTP verify, prefill) runs FlyDSL `pa_decode` (`pa_fwd_asm` does not support gfx1250), and dense layers use it with
  30. · Pautan bukti · GitHub search · MiniMax · pulls · ROCm/aiter · @RElbers
  31. · Pautan bukti · GitHub search · MiniMax · pulls · ROCm/ATOM · @yhl-amd
    Petikan
    under `multi`, which is the prefill-side setup for MiniMax-M3 PD with UMBP:
  32. · Pautan bukti · GitHub search · MiniMax · pulls · ROCm/ATOM · @rbrugaro-amd
    Petikan
    does for the target (vLLM already did this). MiniMax-M3's DSpark draft now validates this config instead of building its own, and declares `draft_skips_online_quant` so its
  33. · Pautan bukti · GitHub search · MiniMax · pulls · Quarkgluonmixture/quarkspace · @github-actions[bot]
    Petikan
    - `intelligence` — minimax-m3 (default) intelligence: catalog 45.4, archive 18.9
  34. · Pautan bukti · GitHub search · MiniMax · pulls · Portkey-AI/models · @polease
  35. · Pautan bukti · GitHub search · MiniMax · pulls · Pinvou/pinvou-agent · @asto18089
  36. · Pautan bukti · GitHub search · MiniMax · pulls · Pb314314/vllm · @Pb314314
    Petikan
    use sparse MLA (dense MLA, DeepSeek-V4 / V4.1, MiniMax-M3, non-MLA models) are not affected; the reason for each is in "Per-model check".
  37. · Pautan bukti · GitHub search · MiniMax · pulls · NVIDIA/TensorRT-LLM · @ziyixiong-nv
    Petikan
    sampler, speculative and QSA tests; the MXFP8 and MiniMax-M3 dispatch tests; GSM8K with DSpark (TP4);
  38. · Pautan bukti · GitHub search · MiniMax · pulls · NVIDIA-NeMo/Megatron-Bridge · @pruprakash
    Petikan
    seeing `TEDotProductAttention`; GREEN 4/4, full file 86 passed, Gemma4/MiniMax-M3/Kimi-K2.5-VL/GPT-OSS unit tests 100 passed. E2E: gpt-oss-20b card inference (TP2 PP2 EP2, `--attention-backend local`) on this branch
  39. · Pautan bukti · GitHub search · MiniMax · pulls · EvolvingLMMs-Lab/lmms-eval · @sahirema
    Petikan
    sample by frame rate ignore it, e.g. vLLM's MiniMax-M3 and Qwen3-VL loaders. On those, `max_frames_num` has no effect: long clips are sampled at the loader's
  40. · Pautan bukti · GitHub search · MiniMax · pulls · overnightworks/speech · @dependabot[bot]
  41. · Pautan bukti · GitHub search · MiniMax · pulls · maximhq/bifrost · @jimaldon
    Petikan
    model catalog datasheet already carries `tencent/deepseek-v4-pro`, `tencent/deepseek-v4-flash` and `tencent/minimax-m3`; live-only models are listed with blank pricing and can be priced via overrides.
  42. · Pautan bukti · GitHub search · MiniMax · pulls · maximhq/bifrost · @octo-patch
    Petikan
    API. New text examples and integration defaults select `MiniMax-M3.1-Flash`, with `MiniMax-M3` as the other current choice. Default management model suggestions put those models first and
  43. · Pautan bukti · GitHub search · MiniMax · pulls · iv-org/invidious · @Fijxu
  44. · Pautan bukti · GitHub search · MiniMax · pulls · iv-org/invidious · @Fijxu
  45. · Pautan bukti · GitHub search · MiniMax · pulls · iv-org/invidious · @Fijxu
  46. · Pautan bukti · GitHub search · MiniMax · pulls · cbc02009/container-images · @mahoro-ai
    Petikan
    `/v1/chat/completions` request to a messages-protocol Go model (e.g. `opencode_go/minimax-m3`, `qwen3.8-max`, `claude-haiku-5-5`) dies mid-stream with `TypeError: 'NoneType' object is not an iterator` → `APIConnectionError` 500 →
  47. · Pautan bukti · GitHub search · MiniMax · pulls · MrGeva/vllm · @MrGeva
    Petikan
    320 questions, concurrency 32, same node, together with MiniMax-M3 model changes):
  48. · Entri katalog · GitHub · MiniMax-AI/minimax-code · code
    Petikan
    afety margin. */ export function computeCompactionTriggerAt(input: { modelId?: string; contextWindow: number; perTurnMaxTokens: number; reserveTokens: number; safetyMarginTokens: number; }): number { if ( input.modelId === 'MiniMax-M3' && (input.contextWindow === 512_000 || input.contextWindow === 1_000_000) ) { return Math.floor(input.contextWindow * 0.9); } const perTurnMaxTokens = Number.isFinite(input.perTurnMaxTokens) && input.perTurnMaxTokens > 0 ? input.perTurnMaxTokens : 0; const reserve = Math.max(input.reserveTokens, perTurnMaxTokens + input.safetyMarginTokens); return Math.max(1, input.contextWindow - reserve); }

    Rujukan dalam kod ini tidak mengesahkan pengecam model yang digunakan.

Spesifikasi

MiniMax M3 ialah model daripada MiniMax dengan pemberat terbuka, dikeluarkan pada 2026-06-01. Tetingkap konteks ialah 1,000,000 token.

Tetingkap konteks
1,000,000
Input
imej · teks · video
Output
teks
Pemberat
Terbuka

Soalan

Apakah tetingkap konteks MiniMax M3?

Tetingkap konteks ialah 1,000,000 token.

Adakah pemberat MiniMax M3 tersedia untuk umum?

Ya. MiniMax menerbitkan pemberatnya.

Bilakah MiniMax M3 dikeluarkan?

Pada 2026-06-01.