Topic
Who runs a benchmark, in what configuration, and whether results are independently reproducible.
GLM-5.3-Flash leads agentic terminal work, DeepSeek V4 Flash is billed as the cheapest per token, and a 2.52B MiniCPM5-2B runs locally under Apache 2.0. The comparison flags most of those numbers as vendor-reported.
Reality
A dev.to post dates WebAssembly 3.0 to September 17, 2025 and credits WasmGC, a 16GB Memory64 heap and advanced exception handling. The figure that decides a client-side build is the job's wall-clock time on the weakest device a team supports.
Reality
General capability barely moves in the rest of Anthropic's table, so the thing teams have to configure is a five-level effort dial that ran one SVG prompt from ten cents to $3.30. The per-token rate never changes.
Reality
Granite Speech 5.0 Turbo CTC drops the language model and three quarters of its output steps. The speed looks real; the accuracy figures are IBM's own.
Reality
Zhipu says GLM-5.3 edged Anthropic and OpenAI on one security benchmark. On the harder exploitation test the gap runs the other way, by 23.6 points.
Reality