Cursor Bench: Grok 4.6 scores 70.8% at a quarter of the leader's price
ChrisGPT · x · 2026-09-02
On the latest Cursor Bench, Fable 5.1 (max) leads the pack, with Grok 4.6 extra high a close second at 70.8% while being 4x cheaper. GPT 5.6 sol max scores 67.2% at half of Fable 5.1's cost. The poster notes Astra and Grok 4.7 are expected to release very soon, which could reshuffle the leaderboard.
More from coding & agent
- Docker sending hundreds of MB before build? Use .dockerignore — _jaydeepkarale · 2026-09-02
- Optimizing 100K Token Skill Descriptions into MCP-Based ARD Server — dSebastien · 2026-09-02
- Perplexity uses Fable 5.1 as orchestrator with GPT 5.6 as cost-efficient subagents — AravSrinivas · 2026-09-02
- Prime Agent 0.9.1 ships with massive performance gains and many bug fixes — samsja19 · 2026-09-02
- Prathmesh Patel discusses agent reliability and testing — jeffiql · 2026-09-02
- Claude + Thrixel build full Roblox games from a single prompt — RanaHanocka · 2026-09-02