Three provider stories set the day. xAI shipped Grok 4.7 at the same price as 4.6. Harvey moved its flagship legal model to Moonshot's Kimi K3 after margins collapsed. Anthropic cleared an early morning error spike before US business hours.
In the news
Grok 4.7 ships with same pricing as 4.6
xAI posted Grok 4.7 on Sep 21. It is live in Cursor, Grok Build and the xAI API. Pricing stays $2 per million input and $6 per million output, the same as Grok 4.6. The launch notes a larger base and longer reinforcement learning. A fast variant runs at double output speed for double price. Author benchmarks show gains on CursorBench 4.0 and internal tests, though the model still trails Fable 5.1 Max on some sets.1
Harvey puts Tenet on Kimi K3 after margin collapse
Bloomberg reported on Sep 21 that Harvey's gross margin fell from about 50 percent to minus 50 percent by June. Token use rose twentyfold under OpenAI and Anthropic usage based pricing. The firm launched Tenet in August 2026. It is a post trained model built on Moonshot's open weight Kimi K3 with Fireworks. The move restored positive margins. Bloomberg names Abridge, Decagon and Ramp as pursuing similar open weight pivots.1
Anthropic error spike resolved early Sep 22
Anthropic opened incident 7g1qpkyz5gxh at 00:57 UTC on Sep 22. It covered elevated errors on Fable 5/5.1, Mythos 5/5.1 and Opus 5. Fable and Mythos recovered by 01:35 UTC. The incident was marked resolved at 02:35 UTC. Impact was 00:50 to 02:10 UTC.1
