MiniMax and Fireworks to host SF fireside on building M3.1 Flash for long-running agents
MiniMax_AI · x · 2026-11-07
MiniMax and Fireworks announced a Code & Cocktails event in San Francisco on November 10, celebrating M3.1 Flash going live on Fireworks. The fireside chat features Sinan Ozdemir (Fireworks) and Victor Su-Ortiz (MiniMax) discussing how the open-weights model was built for long-running multi-step agent workflows, balancing coding, tool use and visual reasoning, and how cheaper cached tokens change agent loop economics.
Event details list M3.1 Flash benchmarks: 62.8 on AutomationBench-AA, 73.7% on DeepSWE v1.1, 91 on BabyVision; priced at $0.15 input / $0.60 output / $0.015 cached per 1M tokens — half the price of M3.
Related event: MiniMax M3.1 Flash Launches on Fireworks with SF Meetup(2 posts)→
More from Companies & People
- Complex Systems podcast dissects Anthropic's S-1: $900B compute commitments and founder control — TheZvi · 2026-11-07
- 'Avatar' VFX House Weta FX Chief: No AI Strategy, No Hollywood Job — nptacek · 2026-11-07
- Analyst predicts China AI consolidation: DeepSeek talent drain leads to 'two big, two small' endgame — teortaxesTex · 2026-11-07
- Google exec: the race for best AI is irrelevant, scaling affordable access is what matters — docmilanfar · 2026-11-07
- Gary Marcus on CNBC: OpenAI is 'possibly the WeWork of AI' — GaryMarcus · 2026-11-07
- Richard Socher's Recursive hiring kernel, platform and DevRel engineers in SF and London — jeffclune · 2026-11-07