artificialwatch
model radar / minimax / minimax-m4

MiniMax M4

MiniMax · the next M, promised for H2
new-gen M promised for H2 · 42 days left on nov 19 m3.1 flash live since oct 19 — open weights 1M context · native multimodal
An email when MiniMax M4 answers on a public API, about 15 minutes after members, plus every frontier launch · no card

Not out yet. MiniMax hasn't released MiniMax M4. MiniMax's M3 landed June 1 with frontier-class coding, a 1M-token context window and native image + video understanding. Its newest M model is M3.1 Flash, rolled out October 19 at $0.15 / $0.60 per 1M tokens with open weights within a week, and MiniMax says an M3.1 Pro will follow. The company has publicly committed to a new M generation in the second half of 2026, and 42 days of that window were left on Nov 19. M3.5 or M4, the watch is the same: the second it answers, ArtificialWatch fires.

✳ the wire

Leaks & news, newest first.

Nov 4 · status

57 days left in H2, and still no M4

MiniMax’s promise of a new M-series generation in H2 2026 has 57 days to run. Its newest M models are M3.1 Flash (Oct 19) and an M3.1 Pro that its launch post says will arrive later. No m4 or m3.5 id is in OpenRouter’s catalog, and MiniMax’s newest Hugging Face repository is M3.1 Flash. via OpenRouter · Hugging Face · MiniMax

Oct 23 · weights

M3.1 Flash weights land on Hugging Face

Four days after launch, the first files of MiniMaxAI/MiniMax-M3.1-Flash went up: 220B parameters by the safetensors count, not gated, under MiniMax’s community license. At launch MiniMax had said only that the weights would be released later. via Hugging Face

Oct 19 · launch

M3.1 Flash ships at $0.15 / $0.60, and Space Bunny was an early version

MiniMax rolled out M3.1 Flash on its API, its M Plan subscription and MiniMax Code: $0.15 per 1M input tokens and $0.60 output, half of M3’s rate, with a 1M-token context. MiniMax says agentic tasks cost 70% less than on its predecessor, and its launch post names the Space Bunny stealth model as an early version. via MiniMax

Sep 23 · leak

Space Bunny, an anonymous free model, appears on OpenRouter and OpenCode

Our sweep logged space-bunny-alpha on OpenRouter and space-bunny-free on OpenCode at 15:12 UTC, both free with a 1M-token context and no maker named. MiniMax claimed it on October 19. via OpenRouter · models.dev

Aug 26 · status

Half-year results: H3 is out, no date for the next M

MiniMax’s first-half 2026 results cite M3 as the period’s core model upgrade and H3, released shortly after, with open weights. The report gives no date for a new M-series model. via MiniMax

Jul 31 · launch

H3 ships — the Hailuo half of the H2 promise

MiniMax released H3, the successor to its Hailuo 01 and 02 video models: an open general-purpose multimodal model that generates video. The M-series half of the H2 commitment has not shipped. via MiniMax

Jul 7 · status

M4 is unannounced — but the successor is already promised

MiniMax has publicly committed to launching a new generation of M-series and Hailuo models in H2 2026. Whether it's badged M3.5 or M4, this page fires either way.

Jun 7 · weights

M3 weights land on Hugging Face

Six days after announcement, the full M3 weights went up — 428B parameters, open, with the community immediately probing what it takes to run at home. via Hugging Face · RunAIHome

Jun 1 · launch

M3: the first open model with frontier coding, 1M context and native multimodality

MiniMax shipped M3 with a 59.0% SWE-Bench Pro score — above GPT-5.5 and Gemini 3.1 Pro — plus a 1M-token window and native image/video understanding, priced at a fraction of the frontier rate cards. via MiniMax · Fireworks

✳ from minimax

The official word.

"A new generation of M-series and Hailuo models in the second half of 2026."

MiniMax, company roadmap · 2026

M3.1 FlashLive since Oct 19 — $0.15 / $0.60, 1M context, open weights (first files Oct 23). A Pro is promised for later.
M3.5Unannounced. M3.1 came first, and MiniMax has named no M3.5.
M4The next full generation this page watches — H2, per the company.
✳ current model in this line

Until MiniMax M4 ships: MiniMax M3.

Which model is current until MiniMax M4 ships?

The current model in this line is MiniMax M3 from MiniMax, live since May 31, 2026.

$0.30 / $1.20 per 1M tokens on MiniMax's API.

Hear it the second the next M answers.

We sweep the public catalogs (OpenRouter, models.dev) every 60 seconds — M3.5 or M4, whichever ships, this watch fires. SMS and a phone call the moment it lands on paid tiers; free email + Chrome push about 15 minutes later.

On the board. You'll hear the second the next MiniMax answers.