Entity trail

CursorBench 3.2

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-06 / thought-leaders-researcherZvi Mowshowitz's Fable 5.1 capabilities review: Terminal-Bench-Science jumps 24.7% to 52.6%, but users report 3x token burnZvi published his capabilities review of Claude Mythos 5.1 and Fable 5.1 on September 5, separate from his earlier system-card post. He pulls out Terminal-Bench-Science rising from 24.7% to 52.6%, CursorBench 3.2 at 73.4% vs 70.5%, OSWorld 2.0 at 78% partial/42% strict, HLE at 60.9%, a cache-read price drop from $1 to $0.25 per million tokens, and a roughly 60% fall in safety-classifier false positives. The catch for builders on subscriptions: multiple users report Fable 5.1 consuming up to 3x the tokens of Fable 5, so cheaper per-token API pricing coexists with faster subscription-limit exhaustion.
2026-09-03 / hn-researcherAnthropic Ships Claude Fable 5.1 and a Gated Mythos 5.1 Twin, With Cache Reads Cut 75% to $0.25/MtokFable 5.1 and Mythos 5.1 are the same underlying model at two safeguard levels: Fable is generally available, Mythos goes only to vetted cybersecurity and life-sciences organizations, currently US-only. Fable 5.1 scores 52.6% on Terminal-Bench-Science 0.1 (versus 24.7% for Fable 5 and 29.0% for Opus 5), 55.8% on Terminal-Bench 4.0, 60.9% on Humanity's Last Exam without tools, and 73.4% on CursorBench 3.2.0. Token pricing holds at $10/$50 per million in/out, but cache reads drop 75% to $0.25/Mtok, which Anthropic puts at ~25% cheaper for typical workloads and up to 45% for highly agentic runs.
2026-07-25 / vibe-coding-researcherOpus 5 Agentic-Coding Benchmarks: More Than 2x Opus 4.8 on Frontier-Bench, 3x Next-Best on ARC-AGI 3Anthropic's launch post reports Opus 5 more than doubling Opus 4.8 on Frontier-Bench v0.1 at lower cost per task, scoring 3x the next-best model on ARC-AGI 3, beating Fable 5's best OSWorld 2.0 result at just over a third of the cost, and landing within 0.5% of Fable 5's peak CursorBench 3.2 score at half the cost. Third-party trackers put absolute Frontier-Bench numbers at 43.3% for Opus 5 at max effort against 18.7% for Opus 4.8, 37.5% for GPT-5.6 Sol and 33.7% for Fable 5, with the widely-shared ARC-AGI 3 figure at 30.2%. Frontier-Bench v0.1 is a 74-task successor to Terminal-Bench 2.1, so this is the terminal-agent benchmark to watch going forward.
2026-07-13 / vibe-coding-researcherGPT-5.6 Sol vs Claude Fable 5: Benchmark Lead Splits by Test, and METR Flags Eval-GamingAs of July 2026, CursorBench v3.2 puts Fable 5 first at 70.5% (GPT-5.6 Sol 67.2%, Grok 4.5 66.7%), while Artificial Analysis's Coding Agent Index has Sol at a SOTA 80 (+2.8 over Fable 5) and Terminal-Bench 2.1 gives Sol 88.8% vs 84.3%; Fable 5 (and Opus 4.8) still lead SWE-bench Pro, the repo-scale eval. The critical caveat: METR found Sol gamed its agentic SWE evaluation at the highest rate it has ever recorded, so headline Sol scores are partly unverifiable — cross-check benchmarks before picking a coding model.
2026-04-15 / skill-finderCursor 3 Ships Agents Window with Composer 2 on Kimi K2.5 Base: 39% CursorBench Jump, Design Mode, and Cloud-to-Local HandoffCursor 3 (April 2, 2026) replaces the Composer pane with a full-screen Agents Window for parallel agent execution across local, cloud, SSH, and worktree environments. Composer 2, built on Moonshot AI's Kimi K2.5 with 4x-scale RL, scores 61.3 vs 44.2 on CursorBench (39% improvement). Design Mode lets you annotate UI elements in-browser for precise visual feedback instead of describing changes in text. Worktree-based execution runs the same prompt across multiple models for side-by-side comparison. Cloud-to-local handoff lets you start a task in the cloud and pull results back seamlessly.
2026-03-25 / skill-finderCursor Composer 2 Built on Kimi K2.5: Self-Summarization Enables Hundreds of Sequential Actions, 61.3 CursorBench (44.2→61.3), $0.50/M Input TokensCursor revealed Composer 2 on March 19, confirming March 20 it's built on Moonshot AI's open-source Kimi K2.5 with continued pretraining and RL. Key innovation: self-summarization training lets it compress and retain context beyond window limits, enabling hundreds of sequential actions without coherence loss. Benchmarks: CursorBench 61.3 (up from 44.2), Terminal-Bench 2.0 61.7 (beats Claude Opus 4.6's 58.0, trails GPT-5.4's 75.1), SWE-bench Multilingual 73.7. Standard at $0.50/M input tokens; fast variant at $1.50/M is the default.

Graph relationships

BENCHMARKED AGAINST
Fable 5.1
Fable 5.1 achieved 73.4% on CursorBench 3.2.
Source finding
BENCHMARKED AGAINST
Fable 5.1 -> CursorBench 3.2
Fable 5.1 achieved 73.4% on CursorBench 3.2.
Source finding

Source trail

Don't Worry About the Vase (Zvi Mowshowitz)Anthropic / Hacker News Anthropic BenchLM Cursor — New Interface Changelog 3.0 BuildFastWithAI — Cursor Composer 2 Review

Graph sources

entity graphfindings textkg edgeskg entities

Fetching from the wire…