Gemini 3.7

1. Mushroom hunting with LLMs: what can go wrong? (quesma.com via hn)

   +1 2w glmgpt-5gemini

   Mushroom identification with AI: GPT-5.6-Sol, Gemini 3.7 Flash, GLM-5.3-Flash and Claude Fable 5.1 benchmarked on poisonous and edible species of FungiTastic. A lot of dangerous errors.

2. Benchmark notes: Fable 5.1 reaches 90/98, with a significant jump in visual performance (www.reddit.com via reddit)

   2w gpt-5sonnetgemini+1

   I ran Claude Fable 5.1 on the current 98-task MindTrial set with the same Python executor available as in the earlier Fable 5, Opus 5 and Sonnet 5 runs. The result was stronger than I expected: 90/98, which is currently the highest raw pas…

3. Celeris-1 Magnus: Fast hybrid diffusion model for agentic work (celeris.ai via hn)

   +5 2w gpt-5geminiagentic

   Magnus is built for agents that need to think, use tools, and get things done, without waiting around. All 97 tasks of τ³-bench banking, head-to-head against gpt-5.6-sol, gpt-5.6-luna and gemini-3.7-flash, reasoning effort as published.

4. Gemini 3.7 flash on cursor cli just deleted my C drive ( via reddit)

   2w geminicursor

   could not extract summary

← all threads