Skip to main content
benchlm.ai

SWE-bench Pro

A stronger coding-agent benchmark than SWE-bench Verified, intended to differentiate frontier models on realistic software engineering work.

Top models on SWE-bench Pro — June 2, 2026

As of June 2, 2026, Claude Mythos Preview leads the SWE-bench Pro leaderboard with 77.8% , followed by Claude Opus 4.8 (69.2%) and Claude Opus 4.7 (Adaptive) (64.3%).

1Closed

Claude Mythos Preview

Anthropic

77.8%
Overall 99Context 1M
2Closed

Claude Opus 4.8

Anthropic

69.2%
Overall 95Context 1M
3Closed

Claude Opus 4.7 (Adaptive)

Anthropic

64.3%
Overall 85Context 1M
35 modelsCoding23% of category scoreCurrentUpdated June 2, 2026

According to BenchLM.ai, Claude Mythos Preview leads the SWE-bench Pro benchmark with a score of 77.8%, followed by Claude Opus 4.8 (69.2%) and Claude Opus 4.7 (Adaptive) (64.3%). There is significant spread across the leaderboard, making this benchmark effective at differentiating model capabilities.

35 models have been evaluated on SWE-bench Pro. The benchmark falls in the Coding category. This category carries a 20% weight in BenchLM.ai's overall scoring system. Within that category, SWE-bench Pro contributes 23% of the category score, so strong performance here directly affects a model's overall ranking.

About SWE-bench Pro

Year

2026

Tasks

Real-world software engineering

Format

Repository task completion

Difficulty

Frontier coding agent

SWE-bench Pro is the more relevant frontier signal when selecting coding agents in 2026. It reflects more realistic difficulty than the older verified subset.

Why we no longer evaluate SWE-bench Verified

BenchLM freshness & provenance

Version

SWE-bench Pro 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

Current

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Leaderboard (35 models)

1
Claude Mythos Preview
77.8%
2
Claude Opus 4.8
69.2%
3
Claude Opus 4.7 (Adaptive)
64.3%
4
Qwen3.7 Max
60.6%
5
MiniMax M3
59%
6
GPT-5.5
58.6%
7
Kimi K2.6
58.6%
8
GLM-5.1
58.4%
9
GPT-5.4
57.7%
10
Qwen 3.6 Max (preview)
57.3%
11
MiMo-V2.5-Pro
57.2%
12
Claude Opus 4.5
57.1%
13
GPT-5.3 Codex
56.8%
14
Qwen3.6 Plus
56.6%
15
Step 3.7 Flash
56.3%
16
MiniMax M2.7
56.2%
17
MiMo-V2.5
56.1%
18
GPT-5.2
55.6%
19
DeepSeek V4 Pro (Max)
55.4%
20
Gemini 3.5 Flash
55.1%
21
GLM-5
55.1%
22
DeepSeek V4 Pro (High)
54.4%
23
Qwen3.6-27B
53.5%
24
Claude Opus 4.6
53.4%
25
DeepSeek V4 Flash (Max)
52.6%
26
Muse Spark
52.4%
27
DeepSeek V4 Flash (High)
52.3%
28
DeepSeek V4 Pro
52.1%
29
Grok 4.20
51.8%
30
Qwen3.5 397B
50.9%
31
Kimi K2.5
50.7%
32
Qwen3.6-35B-A3B
49.5%
33
Laguna M.1
49.2%
34
DeepSeek V4 Flash
49.1%
35
Laguna XS.2
46.3%

FAQ

What does SWE-bench Pro measure?

A stronger coding-agent benchmark than SWE-bench Verified, intended to differentiate frontier models on realistic software engineering work.

Which model scores highest on SWE-bench Pro?

Claude Mythos Preview by Anthropic currently leads with a score of 77.8% on SWE-bench Pro.

How many models are evaluated on SWE-bench Pro?

35 AI models have been evaluated on SWE-bench Pro on BenchLM.

Compare Top Models on SWE-bench Pro

Claude Mythos Preview vs Claude Opus 4.8Claude Opus 4.8 vs Claude Opus 4.7 (Adaptive)Claude Opus 4.7 (Adaptive) vs Qwen3.7 MaxQwen3.7 Max vs MiniMax M3
Last updated: June 2, 2026 · BenchLM version SWE-bench Pro 2026

The AI models change fast. We track them for you.

For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.

Free. No spam. Unsubscribe anytime.