how tokenspeed made 397b qwen3.5 run at 580 tps?
microsoft mai-image-2.5 hits #3 on arena – first time microsoft breaks top 5
xiaomi follows deepseek's playbook: mimo-v2.5-pro api now matches deepseek-v4-pro pricing to the cent
china's ai market: capital, talent, and the closing gate
openrouter raised $113m at $1.3b valuation. here's where the money is going
minimax m3: sparse attention hits 15.6x decoding speedup at 1m tokens
hermes cut startup time 63% and just beat codex on speed
hermes is now officially faster than codex. here's how they did it
grok foundation model v9-medium training completed – looks like it's grok 4.5
google's agent cracked 9 erdos problems, 44 oeis conjectures