Haiku 5.5 knocks Jev off the WebMCP top spot
Idan Levin's WebMCP benchmark has a new winner, and it is not Jev: Haiku 5.5 took first on the combined success, speed and cost score. On the same day OpenRouter shipped Armor, a tool call guard built on Jev 1.13.
Built with it
Moritz Kremb open sources Eyes, a video preprocessor for agents
Moritz Kremb released Eyes, a free local tool that processes a video once and hands an agent labelled frame sheets, every spoken word with a timestamp, and on-screen text with where and when it appears. His argument is that Claude otherwise samples a few frames, guesses the gaps, and starts over on follow-up questions, burning time and tokens.
Someone measured it
OpenRouter ships Armor, a tool call guard running on Jev 1.13
OpenRouter launched Armor in beta, which scores each tool call against a rule set plus a slice of the transcript before the agent executes it.
regex-based filtering often isn't enough to detect unwanted tool calls
The case against
Idan Levin's WebMCP benchmark puts Haiku 5.5 ahead of Jev
Idan Levin added six models to the WebMCP benchmark and says Haiku 5.5 using WebMCP now takes first on the combined success, speed and cost score, ahead of GPT-6 Luna and previous winner Jev.
haiku 5.5 is ~3.5 seconds per task, which is more than 2x faster than other models with webmcp