Fetching from the wire…
Models2026-09-22 · source-backed
At Apsara in Hangzhou on September 22, CEO Eddie Wu announced a T-Head-designed chip claimed at 3x the M890's throughput, scaling to clusters of 500,000 units, mass production targeted for Q1 2027. Wu said Qwen 4 is already training and the 4.5 and 5 series will run two to four times the current flagship's ~2.4 trillion parameters. The pairing matters more than either half: Alibaba is claiming it can train frontier-scale models on domestic silicon. (Crypto Briefing)
Each link below shares sources, entities, or timing with this story.
Alibaba's model was reported best overall on Artificial Analysis' Agentic Index, drawing 540 points on HN. Readers watching the page saw Qwen at 55.4 vs Opus Max at 55.3, then on reload Opus Max at 59.2 vs Qwen at 58.4. George from Artificial Analysis replied in-thread that th...
Ollama cut v0.34.0-rc1 on September 5 at 23:49 UTC, and the headline item changes the shape of the local-versus-hosted decision rather than the performance of either side: Ollama-hosted open models can be selected directly inside ChatGPT Desktop, with setup driven from the Oll...
The Max tier is Qwen's flagship proprietary line, distinct from the open-weight Qwen3 series, continuing the Chinese frontier release cadence alongside Moonshot's K3. The practical question for anyone outside China is whether Max-tier access lands on international API endpoint...
Alibaba's Tongyi Lab shipped Qwen-RobotSuite on June 15: Qwen-RobotNav for navigation, Qwen-RobotManip for grasping and manipulation, and Qwen-RobotWorld, a world model predicting future physical states from observations plus natural-language actions. Already in enterprise pil...
Junyang Lin (tech lead who built Qwen from lab project to 600M+ downloads) and Yu Bowen (post-training head) resigned one day after Qwen 3.5 launched. Huibin (Qwen Code lead) had already left for Meta in January. The catalyst: Alibaba dismantled Lin's vertically-integrated R&D...
Alibaba open-sourced it September 20, unifying text-to-image, editing and native transparent generation in one model: 7B across 32 single-stream DiT layers, paired with a Qwen3-VL 8B text encoder and a 64-channel RGBA VAE at 16x spatial compression. It scores 60.28 on the publ...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.