Gemma 4 31B FP8 Holds Its Own Against Claude Sonnet 4.6 in BenchLM Tests
BenchLM’s head‑to‑head benchmark pits Google’s Gemma 4 31B (FP8) against Anthropic’s Claude Sonnet 4.6. While Sonnet scores higher overall (83 vs 64) and leads on coding, Gemma narrows the gap on multimodal and knowledge tasks and offers free token pricing plus a 256K context window.