PROXIMAL

Proximal is a research
lab focused on data

Read Blog
Introducing FrontierSWE

Our Ultra-Long-Horizon Coding Benchmark

See Benchmark→

Leaderboard

#ModelAVG RANKDominance
1
Claude Fable 5
Claude Code
2.4789%
2
Grok 4.5
Grok CLI
4.0978%
3
Claude Opus 4.8
Claude Code
4.8273%
4
GLM-5.2
Claude Code
4.8572%
5
GPT-5.5
Codex
5.2170%
6
Claude Opus 4.7
Claude Code
6.4761%
7
Claude Opus 4.6
Claude Code
7.5953%
8
GPT-5.4
Codex
7.8851%
9
Composer 2.5
Cursor CLI
9.6538%
10
Gemini 3.1 Pro
Gemini CLI
9.7937%
11
GLM-5.1
Claude Code
11.0029%
12
DeepSeek V4 Pro
Claude Code
11.1827%
13
Kimi K2.6
Kimi CLI
11.4425%
14
Kimi K2.5
Kimi CLI
11.5025%
15
Qwen3.6-Plus
Qwen Code
12.0621%
Rank: avg position across tasks (lower = better)Dominance: win rate vs random opponent on task.
Latest
Announcements, benchmark releases, and work that shows the direction of the lab.

Our Problems

Some of the problems we are working on.

FrontierSWE

Benchmarking coding agents at the limits of human abilities.

Announcing Proximal

A new type of data company.