Fetching from the wire…
01
02
03
04
05
06
07
08
Agents2026-08-21 · source-backed
BenchLM's leaderboard ranks 36 models on interactive tool-calling over MCP at an advanced tier, with Kimi K3 third at 84.2%. BenchLM.ai It's display-only and excluded from composite rankings, refreshing quarterly, so treat the spread as directional. The interesting thing isn't the ordering, it's that a four-point band separates the top three and there's now a public number for the thing everybody actually uses models for.
Each link below shares sources, entities, or timing with this story.
Cursor uses MCP / Shared entities / Earlier coverage
Linked by a graph relationship (Cursor uses MCP); both cover Kimi K3, MCP, Opus; earlier Kimi K3 coverage from 2026-07-26.
Anthropic released MCP / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Anthropic released MCP); both cover Muse Spark, Opus; overlapping topics (kimi, model).
Claude Code uses MCP / Shared entities / Earlier coverage / Tension
Linked by a graph relationship (Claude Code uses MCP); both cover MCP, Opus; earlier MCP coverage from 2026-08-05.
Anthropic released MCP / Shared entity: BenchLM / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (Anthropic released MCP); both cover BenchLM; reported by the same outlet (benchlm.ai).
Claude Code uses MCP / Shared entities / Earlier coverage / Tension
Linked by a graph relationship (Claude Code uses MCP); both cover MCP, Opus; earlier MCP coverage from 2026-05-17.
Anthropic released MCP / Shared entity: Opus / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Anthropic released MCP); both cover Opus; overlapping topics (actually, model).
Linked by a graph relationship (Anthropic released MCP); both cover Opus; overlapping topics (actually, model).
OpenAI supports MCP / Shared entities / Earlier coverage
Linked by a graph relationship (OpenAI supports MCP); both cover Kimi K3, Opus; earlier Kimi K3 coverage from 2026-08-15.