Up to date
MiniMaxAI MiniMax Model License (minimax-community) out 8 days ago
220B parameters, images and text in, text out, 977K tokens of context. 2 providers serve it from $0.15 per million input tokens.
Verified 2026-10-25 Rank #432 in LLMs · score 78.5 method v1 Sources 4 Downloads 30d 5.8k · #534 of 619 tracked
Changed this week: MiniMax M3.1 Flash entered the release radar as served, open-weights (Hugging Face, 2026-10-23).2026-10-23
Picked by rule from 6 tracked pieces: the lab's own post, the most shared, a field report, the latest.
MiniMax Research officially rolls out MiniMax M3.1 Flash, a next-generation model designed for agentic tasks, coding, and cybersecurity.
MiniMax-M3.1-Flash is a proprietary multimodal model released on October 19, 2026, that scores 43 on the Artificial Analysis Intelligence Index.
MiniMax M3.1 Flash is a native multimodal Mixture-of-Experts model released October 18, 2026, designed for coding, agentic workflows, and computer use.
The 4 picks above, plus these 2. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
| # | Cheapest live row | In /1M | Out /1M | Context | Quant | Seen |
|---|---|---|---|---|---|---|
| 1 | Parasail | $0.15 | $0.60 | — | — | Oct 24 |
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
Served by 2 API providers; 25 quantized builds for local use.
Weights only, an estimate: leave room for the KV cache. Method
About 132 GB at INT4, 265 GB at FP8, 529 GB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2026-10-17), amber triangles are 6 dated articles (each links to it).
30 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
| Pipeline tag | image-text-to-text |
|---|---|
| Library | transformers |
| Architectures | MiniMaxM3SparseForConditionalGeneration |
| Model type | minimax_m3_vl |
| Params by dtype | BF16: 3,323,222,528 · F32: 46,339,200 · F8_E4M3: 10,560,208,896 · U8: 206,561,083,392 |
| Size band | XXL |
| MoE | yes |
| Context length | 1,000,000 (provider listing) |
| Papers | arXiv:2606.13392 |
| License | MiniMax Model License (minimax-community) (other) |
| Hub snapshot | 2026-10-25 |
Weights-only memory for MiniMax-M3.1-Flash: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 529 GB · Multi-GPU node (<640 GB) |
|---|---|
| FP8 (1 byte/param) | 265 GB · Multi-GPU node (<640 GB) |
| INT4 (0.5 bytes/param) | 132 GB · Multi-GPU node (<640 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.
quantized: 25finetune: 5
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| Baekpica/MiniMax-M3.1-Flash-VQ-GGUF | quantized | GGUF | — | — | other | Baekpica | 0 | 3.2k |
| webmp3/Sakura-MiniMax-M3.1-Flash-Q3_K_M-GGUF | quantized | GGUF | — | — | other | webmp3 | 0 | 2k |
| LyJonathan/MiniMax-M3.1-Flash-GGUF | quantized | GGUF | — | — | other | LyJonathan | 5 | 1.5k |
| unsloth/MiniMax-M3.1-Flash-GGUF | quantized | GGUF | — | — | other | unsloth | 19 | 589 |
| AtomicChat/MiniMax-M3.1-Flash-GGUF | quantized | GGUF | — | — | other | AtomicChat | 2 | 807 |
| incoai/MiniMax-M3.1-Flash-DFlash2 | finetune | — | 2.12B | M | apache-2.0 | incoai | 12 | 107 |
| unsloth/MiniMax-M3.1-Flash | quantized | — | 220B | XXL | other | unsloth | 7 | 271 |
| mlx-community/MiniMax-M3.1-Flash-Text-4bit | quantized | -4bit | 426B | XXL | other | mlx-community | 4 | 289 |
| mlx-community/MiniMax-M3.1-Flash-oQ4e-mtp | quantized | — | 427B | XXL | other | mlx-community | 3 | 275 |
| amd/MiniMax-M3.1-Flash-Quark-MXFP4 | quantized | MXFP4 | 220B | XXL | other | amd | 1 | 294 |
| mlx-community/MiniMax-M3.1-Flash-oQ3e-mtp | quantized | — | 427B | XXL | other | mlx-community | 2 | 243 |
| mlx-community/MiniMax-M3.1-Flash-oQ3.5e-mtp | quantized | — | 427B | XXL | other | mlx-community | 2 | 175 |
| Mia-AiLab/MiniMax-M3.1-Flash-EXL3-EAQ-v1-3.06-bpw | quantized | EXL3 | 84.3B | XL | — | Mia-AiLab | 3 | 88 |
| webmp3/Sakura-MiniMax-M3.1-Flash-Q2_K-GGUF | quantized | Q2_K | — | — | other | webmp3 | 2 | 94 |
| mlx-community/MiniMax-M3.1-Flash-Text-3bit | quantized | — | 426B | XXL | other | mlx-community | 1 | 135 |
| vcruz305/MiniMax-M3.1-Flash-EXL3-SAGE-2.15bpw | quantized | EXL3 | 58.4B | XL | other | vcruz305 | 2 | 34 |
| Mia-AiLab/MiniMax-M3.1-Flash-EXL3-EAQ-v1-3.5-bpw | quantized | EXL3 | 97.1B | XL | — | Mia-AiLab | 1 | 56 |
| RadixArk/Minimax-M3.1-Flash-Dflash2 | finetune | — | 2.05B | M | — | RadixArk | 1 | 52 |
| webmp3/Sakura-MiniMax-M3.1-Flash-Q4_K_M-GGUF | quantized | GGUF | — | — | other | webmp3 | 1 | 39 |
| primitive-ai/MiniMax-M3.1-Flash-pruned-272B-NVFP4 | quantized | NVFP4 | 143B | XXL | other | primitive-ai | 1 | 34 |
| infosave/MiniMax-M3.1-Flash-cmf | quantized | — | — | — | apache-2.0 | infosave | 1 | 31 |
| BobClawblaw/MiniMax-M3.1-Flash-MLX-mix34 | finetune | MLX | 431B | XXL | other | BobClawblaw | 1 | 29 |
| vcruz305/MiniMax-M3.1-Flash-EXL3-SAGE-4.26bpw | quantized | EXL3 | 118B | XXL | other | vcruz305 | 1 | 18 |
| inferencerlabs/MiniMax-M3.1-Flash-MLX-Q9 | quantized | MLX | — | — | — | inferencerlabs | 0 | 59 |
| webmp3/Sakura-MiniMax-M3.1-Flash-Q5_K_M-GGUF | quantized | GGUF | — | — | other | webmp3 | 0 | 35 |
| BobClawblaw/MiniMax-M3.1-Flash-MLX-mix256k | finetune | MLX | 431B | XXL | other | BobClawblaw | 0 | 14 |
| gatilin/minimax_m3_1_flash_vision | finetune | — | — | — | other | gatilin | 0 | 12 |
| mondk/MiniMax-M3.1-Flash-GGUF | quantized | GGUF | — | — | other | mondk | 0 | 0 |
| tiny-random/minimax-m3.1-flash | quantized | — | 14.0M | S | — | tiny-random | 0 | 0 |
| yujiepan/minimax-m3.1-flash-tiny-random | quantized | — | 14.0M | S | — | yujiepan | 0 | 0 |
Hosted endpoints observed for MiniMax-M3.1-Flash, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.15/M.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| Novita hf | minimax/minimax-m3.1-flash | 977K | — | $0.15 | $0.60 | 66.4 | 1962 ms | yes | staging | 2026-10-25 |
| Parasail parasail | parasail-minimax-m31-flash | — | — | $0.15 | $0.60 | — | — | — | serverless | 2026-10-24 |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
Hub snapshot: 2026-10-25 · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-24 21:52 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
