MiniMax (official)

1.3K posts

MiniMax (official) banner
MiniMax (official)

MiniMax (official)

@MiniMax_AI

Agent: @MiniMaxAgent Token Plan: https://t.co/BDCycxepZw API: https://t.co/fHRdSV7BwZ Discord: https://t.co/uYnFNkYh7y

San Francisco Katılım Ocak 2025
713 Takip Edilen89.3K Takipçiler
Florian Brand
Florian Brand@xeophon·
maybe @natolambert was right and the space needle is more impressive than the golden gate bridge
Florian Brand tweet media
English
5
0
86
7K
MiniMax (official)
MiniMax (official)@MiniMax_AI·
Such a great night🥂 So glad to have you @kenyeung128 and the @Wati_io team for joining! Big thanks to @boltdotnew for making it happen, and to @UnicornerNews for co-hosting alongside us.
Ken Yeung@kenyeung128

Great night at the AI Champions Dinner in SF. Huge thanks to @MiniMax_AI, @Unicorner, and @bolt_new for bringing this group together @EricSimons @tweetsbygarrett ! The highlight of the night was the conversation around how we shift AI from just a product collaboration tool into something that drives genuine human connection. That is exactly how we are building at @wati_io. AI handles the complexity, so you can connect with humans at scale. 📷📷

English
3
1
31
5.5K
MiniMax (official)
MiniMax (official)@MiniMax_AI·
Day 1 at @stripe Sessions was 🔥 Our booth was packed with amazing conversations all day. Thank you for the energy! Day 2 is on. Find us at Booth 604 👋
MiniMax (official) tweet mediaMiniMax (official) tweet mediaMiniMax (official) tweet mediaMiniMax (official) tweet media
English
3
3
307
17.8K
MiniMax (official)
MiniMax (official)@MiniMax_AI·
Excited to power @mira_t_me's AI agent on Telegram — multimodal, at scale, for everyday users. Welcome to the ecosystem 🤝
Mira@mira_t_me

We're partnering with @MiniMax_AI — the core model powering Mira Best price-to-quality ratio we've tested. Built in Shanghai, used by 236M+ users worldwide. Fast, multimodal, and scales. The future of AI agents runs on great infrastructure 🤝

English
8
7
138
15.3K
MiniMax (official) retweetledi
Cloudflare Developers
Cloudflare Developers@CloudflareDev·
Music-2.6 from @MiniMax_AI is free on Cloudflare this week! Generate full-length songs or instrumentals from a text prompt, with optional auto-generated lyrics. Fast global inference on Cloudflare’s network. Perfect for building music apps on Cloudflare Workers. Try it free before the promo ends: developers.cloudflare.com/ai/models/mini…
English
7
24
224
24.6K
MiniMax (official)
MiniMax (official)@MiniMax_AI·
Really excited about this one. Music 2.6 is now available on @Cloudflare AI — full songs with vocals, instrumentals, covers, all from text. We want honest feedback from real users. Give it a spin and let us know what hits (and what doesn't).
Cloudflare@Cloudflare

Music-2.6 from @MiniMax_AI is free on Cloudflare this week! Generate full-length songs or instrumentals from a text prompt, with optional auto-generated lyrics. Fast global inference on Cloudflare’s network. Perfect for building music apps on #CloudflareWorkers. Try it free before the promo ends: developers.cloudflare.com/ai/models/mini…

English
11
14
203
33.5K
MiniMax (official)
MiniMax (official)@MiniMax_AI·
We'll be at @stripe Sessions this week! 🔖 Booth 604 📅 Apr 29–30 📍 Moscone West, SF Come say hi 👋
MiniMax (official) tweet media
English
6
3
300
18.8K
MiniMax (official)
MiniMax (official)@MiniMax_AI·
@SentientAGI @goose_oss Great news! The M2.5 can complete all tasks at only 1/30 of the cost. You may also try M2.7 — it delivers even better performance at the same price point.
English
1
0
32
1.7K
Sentient
Sentient@SentientAGI·
By engineering skills, prompts, and configs with MiniMax M2.5 (@MiniMax_AI) and the Goose (@goose_oss) open-source harness, the top Arena teams hit ~70% accuracy on OfficeQA at $1.74/run — near-frontier performance at 1/30th the cost of Claude Opus 4.5. The takeaway: open-source models aren't just cheaper. With the right harness and prompting, they win significantly on accuracy-per-dollar against closed-source competition. Read more about how Arena Cohort 0 did it, and why harness choice, prompt density, and skills matter more than you think ↓
Sentient tweet media
Sentient@SentientAGI

x.com/i/article/2046…

English
23
11
113
11.7K
MiniMax (official)
MiniMax (official)@MiniMax_AI·
MiniMax Music 2.6 just got a new home 🎵 Sonus AI Music (@sonusmusic89867) is now powered by our latest model. Richer vocals, sharper instrument detail, stronger style expression. Go make something good!
Sonus - AI Music@SonusAI_Music

🎶 Sonus AI Music now integrates MiniMax 2.6. Create with richer vocals, sharper instrument detail, stronger style expression, and smoother overall sound. Start creating now: sonusai.net #Sonus #AIMusic #MiniMax

English
10
9
110
17.3K
Dan-ish
Dan-ish@danish_sol·
Built 2 pipelines, a research and dev lane. Each is a subagent /profile with their own set of files, skills etc. integrated with an obsidian vault for their individual outputs so everything is traceable and documented. I am blown away. Coolest thing ever. #hermes @NousResearch
Dan-ish tweet mediaDan-ish tweet media
English
2
0
1
86
Marc Lindsay
Marc Lindsay@The0neTheOnly·
Yay Me.... Hermes with @MiniMax_AI is now working for me in a Workflow harness. I'll post a video later I now have an E2E workflow management harness that works directly with Hermes @NousResearch that is based on the concepts that @garrytan is sharing. Fat Skills. Scalable, Adaptable, Dependable It is as good as your process, your skills files, your agent files. It takes a vertical skill and makes it horizontal. aka scale..... From the moment I setup Openclaw then hermes the one thing I have always despised..... Cron based heartbeat, it just felt like such a waste and inneficient solution. You want 20 developers in the implementor stage. Done. Zero Heartbeats needed Zero agent cron's needed Zero downtime or wait time The best part.... it's being engineered for ANY process. Any Team workflow Any Project workflow This isn't a development pipeline. It's an everything pipeline
Marc Lindsay tweet media
English
1
0
2
160
MiniMax (official) retweetledi
Ti Guo
Ti Guo@ti_guo_·
Interesting local agent pattern: Hermes Agent (@NousResearch) + orchestrator and sub-agents on different local LLMs. @loktar00's Hermes: MiniMax M2.7 routing (10B active, fast), Qwen 27B sub-agents, tool sandbox with approvals. Closer to capable local agents than people think. (Disclosure: joining @MiniMax_AI soon)
Loktar 🇺🇸@loktar00

Wow I'm 100% fully local now for Hermes. Mini max 2.7 is my orchestrator, Qwen 27b handle sub agents.. this is actually insane. Just had it do a deep vibe research and while it was a tad slower than GLM-5-Turbo it did it without fault. I promise I'm not making $ Minimax 😂

English
10
15
255
29.8K
MiniMax (official) retweetledi
GitLawb
GitLawb@gitlawb·
MiniMax is now official supported provider for OpenClaude! cc @MiniMax_AI
English
14
41
434
42.2K
MiniMax (official)
MiniMax (official)@MiniMax_AI·
unstoppable
stevibe@stevibe

MiniMax M2.7 is 230B params. Can you actually run it at home? I tested Unsloth's UD-IQ3_XXS (80GB) on 4 different rigs: 🟠 4x RTX 4090 (96GB): 71.52 tok/s, TTFT 1045ms 🟢 4x RTX 5090 (128GB): 120.54 tok/s, TTFT 725ms 🟡 1x RTX PRO 6000 (96GB): 118.74 tok/s, TTFT 765ms 🟣 DGX Spark (128GB) — 24.41 tok/s, TTFT 741ms Backend: llama.cpp. Context: 32k. Max tokens: 4096. I went with IQ3_XXS because it's the biggest quant that fits in 96GB VRAM while still leaving safe headroom for 32k context. Same quant across all four rigs, fairest comparison I could run. Now look at rough peak GPU power draw: 🟠 4x4090 → 1,800W peak (450W × 4) 🟢 4x5090 → 2,300W peak (575W × 4) 🟡 RTX PRO 6000 → 600W peak 🟣 DGX Spark → 240W peak (whole system) The RTX PRO 6000 is the quiet winner. One card, 96GB, matching a 4x5090 rig at roughly a quarter of the power and zero multi-GPU headaches. Best tokens-per-watt by a wide margin. DGX Spark is slow on generation but pulls the least power of any rig here, around 240W for the whole system. Prefill-friendly, memory-rich, wall-socket-friendly. And yes, plenty of people cap their cards. Even then, 4x 4090 or 4x 5090 still pulls well over 1,200W from the GPUs alone.

English
16
14
361
42.7K

Keşfet

@xeophon @natolambert @kenyeung128 @Wati_io @boltdotnew @UnicornerNews @stripe @mira_t_me