MiniMax
MiniMax is a Chinese AI foundation-model company founded in December 2021 by Yan Junjie and Zhou Yucong, both former SenseTime executives. Headquartered in Shanghai, it develops the MiniMax-M2 reasoning LLM, M1, the Hailuo-02 text-to-video model, and Speech-02 audio model. Consumer apps include Talkie (international) and Xing Ye (China). MiniMax IPO'd on the Hong Kong Stock Exchange (00100.HK) on 9 January 2026, with shares surging more than 70% on debut and market cap briefly exceeding HK$90 billion (~US$11.5B). Backers include MiHoYo (angel), Alibaba, Tencent, Hillhouse, Sequoia China, and IDG Capital. Its M2.7 model recently demonstrated 400 TPS on SambaNova hardware.
MiniMax, founded by ex-SenseTime executives in December 2021, is rapidly expanding its AI footprint. It has shipped MiniMax M3, MiniMax M2.5, Hailuo-02 text-to-video, Astra Voice 2.0, and Speech-02 audio models. It competes directly with Google, TSMC, Anthropic, OpenAI, and Claude Code—while simultaneously using Claude Code and Hugging Face. This dual dependency and rivalry suggests a strategic posture of borrowing from competitors to build its own stack. Partnerships with Nous Research, Together AI, Nvidia, and Llama signal a deliberate ecosystem play, not a solo sprint. With only 7 mentions in 30 days, MiniMax flies under the radar but ships aggressively. Its MMX-CLI and MaxHermes products extend reach into developer tooling. The question is whether MiniMax can sustain its multi-front competition without a dedicated chip supply chain like TSMC's.
- ·Competes with OpenAI, Anthropic, Google, TSMC—while using Claude Code and Hugging Face.
- ·Shipped 7+ products including M3, M2.5, Hailuo-02, Astra Voice 2.0, Speech-02, MMX-CLI, MaxHermes.
- ·Partners with Nous Research, Together AI, Nvidia, and Llama to build an ecosystem.
- ·Low mention count (7 in 30 days) despite aggressive product velocity.
Signal Radar
Five-axis snapshot of this entity's footprint
Mentions × Lab Attention
Weekly mentions (solid) and average article relevance (dotted)
Timeline
20- Research MilestoneJun 12, 2026
M3 model exceeded human gold-medal threshold on math benchmarks using MaxProof framework.
View source - Research MilestoneJun 1, 2026
Published research paper on MiniMax-M2 series, designed end-to-end for agentic deployment with agent-driven data pipelines
- Product LaunchApr 27, 2026
Made Music-2.6 model available for free on Cloudflare's platform for one week.
View source - Product LaunchApr 23, 2026
Showcased a project where the MiniMax M2.7 model generated a complete 'Flappy Lobster' game from a text prompt.
View source - PartnershipApr 20, 2026
Added as official provider for OpenClaude framework
View source- framework:
- OpenClaude
- role:
- official_provider
- Product LaunchApr 18, 2026
Open-sourced M2.7 model claiming state-of-the-art scores on SWE-Pro (56.22%) and Terminal Bench 2 (57.0%)
View source- benchmark scores:
- SWE-Pro: 56.22%, Terminal Bench 2: 57.0%
- PartnershipApr 16, 2026
Partnered with NousResearch to integrate Hermes framework with MiniMax's M2.7 model.
View source - Product LaunchApr 16, 2026
Launched MaxHermes, a cloud-hosted and managed version of the Hermes agent framework.
View source - PartnershipApr 16, 2026
Partnered with NousResearch to integrate Hermes framework with MiniMax's M2.7 model.
View source - Product LaunchApr 16, 2026
Launched MaxHermes, a cloud-hosted agent framework, in partnership with NousResearch.
View source - Research MilestoneApr 14, 2026
Claimed 26% improvement on BU Bench via social media post without releasing paper or code
View source - Product LaunchApr 13, 2026
Open-sourced three 'Music Skills' for its MMX-CLI agent platform
View source - Product LaunchApr 12, 2026
MiniMax open-sources its M2.7 model, revealing a 'self-evolution' training methodology.
- Product LaunchApr 12, 2026
Made M2.7 model available through NVIDIA's NIM endpoints with OpenClaw/NemoClaw support
View source - Product LaunchApr 11, 2026
Launched MMX-CLI, a command-line interface infrastructure for AI agents.
View source - Product LaunchApr 10, 2026
Launched MMX-CLI, a command-line interface tool for automating multimodal AI agent workflows
- Research MilestoneMar 31, 2026
Released M2.7 AI agent that autonomously rewrites its own operational harness, achieving 9 gold medals on MLE Bench Lite without retraining.
View source - Research MilestoneMar 18, 2026
MiniMax M2.7 ties Gemini 3.1 on MLE Bench Lite machine learning competition trials
View source- benchmark:
- MLE Bench Lite
- result:
- tied
- medal rate:
- 66.6%
Relationships
29Developed
Competes With
Partnered
Uses
Frequently appears with
9Entities that show up in the same articles — shared coverage, not a stated relationship.
Recent Articles
5MiniMax M3 Exceeds Human Gold-Medal on Math Benchmarks via MaxProof
+MiniMax's M3 exceeded human gold-medal on math benchmarks via MaxProof, but no scores or details were disclosed.
99 relevanceMiniMax-M3 Scores 55 on AI Index, Open-Source Lead Looms
+MiniMax-M3 scored 55 on the Artificial Analysis Intelligence Index, set to become the leading open-source model once weights are released.
87 relevanceMiniMax M3: Sparse Attention, 1M Context, Multimodal via Together
+MiniMax M3 uses sparse attention for 1M context and multimodality, with Together AI serving fast inference.
95 relevanceMiniMax Claims 26% BU Bench Gain, Details Scarce
-MiniMax claimed 26% BU Bench improvement without paper or code. Unverifiable claim reduces credibility.
95 relevanceMiniMax M3 Sparse Attention: 15.6x Decoding Speedup at 1M Tokens
+MiniMax M3 sparse attention achieves 9.7x prefilling and 15.6x decoding speedup at 1M tokens, reversing M2's full-attention stance.
100 relevance
Predictions
No predictions linked to this entity.
AI Discoveries
4- observationactive5d ago
Sentiment divergence: MiniMax vs Minimax M3
MiniMax and Minimax M3 have a 'developed' relationship (4 evidence articles) but their recent sentiment has diverged significantly: MiniMax=0.20, Minimax M3=0.57 (gap=0.37). Sentiment divergence between related entities often signals an emerging conflict, leadership change, or strategic shift.
70% confidence - observationactiveJun 5, 2026
Research: Sparse Attention & Long Context [accelerating]
State of art: MiniMax M3 uses sparse attention for 1M context with multimodality; Together AI serves fast inference.. Key insight: Sparse attention is commoditizing long-context inference, enabling memory-as-a-service and agentic workflows.. Leading: MiniMax, Together AI
70% confidence - observationactiveJun 5, 2026
Lifecycle: MiniMax
MiniMax is in 'established' phase (1 mentions/3d, 3/14d, 28 total)
90% confidence - hypothesisactiveFeb 25, 2026
H: MiniMax's surge in visibility precedes a major, Western-focused product launch or API release within
MiniMax's surge in visibility precedes a major, Western-focused product launch or API release within 1 month, aimed directly at competing with OpenAI's and Google's core models, leveraging the 'Sovereign AI' narrative to position itself as a non-US alternative.
75% confidence
Sentiment History
| Week | Avg Sentiment | Mentions |
|---|---|---|
| 2026-W17 | 0.50 | 1 |
| 2026-W18 | 0.45 | 2 |
| 2026-W19 | 0.60 | 1 |
| 2026-W21 | 0.30 | 2 |
| 2026-W22 | 0.70 | 1 |
| 2026-W23 | 0.05 | 2 |
| 2026-W24 | 0.50 | 2 |