FrontierSWE leaderboard
As of November 8, 2026, Gemini 4 Argon (Google DeepMind) leads FrontierSWE at 55%. 6 systems ranked. Every score here is published by the model's own lab; each row links to the page it comes from.
What is FrontierSWE? FrontierSWE gives AI agents ultra-long engineering and research projects, such as reimplementing git in Zig or speeding up a compiler, with up to 20 hours to work. Made by Proximal, September 2026 (v1 in April 2026).
Top models on FrontierSWE, November 2026
| # | Model | Score | Price / 1M | Context | Source |
|---|---|---|---|---|---|
| 1 | Gemini 4 Argon Google DeepMind | as reported | $2 / $10 | 1M | Vendor-reported Google DeepMind |
FrontierSWE under other settings
Labs report FrontierSWE under different settings, and a score from one setting is not comparable with a score from another. Each setting has its own table, ranked on its own.
Version not stated
| # | Model | Score | Price / 1M | Context | Source |
|---|---|---|---|---|---|
| 1 | Kimi K3 Moonshot (Kimi) | max effort | $3 / $15 | 1M | Vendor-reported Moonshot AI |
| 2 | GLM-5.3 Zhipu (Z.ai / GLM) | as reported | $1.4 / $4.4 | 1M | Vendor-reported Z.ai (Zhipu) |
| 3 | Qwen3.8 Max Alibaba (Qwen) | as reported | $2 / $6 | 1M | Vendor-reported Alibaba (Qwen) |
| 4 | GLM-5.2 Zhipu (Z.ai / GLM) | as reported | $1.4 / $4.4 | 1M | Vendor-reported Z.ai (Zhipu) |
| 5 | Qwen3.7 Max Alibaba (Qwen) | as reported | $2.5 / $7.5 | 1M | Vendor-reported Alibaba (Qwen) |
Can you trust FrontierSWE scores?
No independent leaderboard publishes FrontierSWE results for these models yet, so every number on this page is the lab's own report of its own model. Labs choose the harness, the effort level and the subset they report, and they rarely publish failures. Read the setting under each score before comparing two rows.
- Runs use Proximal's own harness, which keeps the agent aware of remaining time and grading; models in their native harnesses are not evaluated.
- v2 added 21 tasks, retired four and changed scoring, so v1 and v2 are not comparable.
- Some entries fall back to another model when a content filter blocks a task, so not every row is a single-model run.
How FrontierSWE works
The task
The agent works in a sandbox with a self-check tool on one long project, then a hidden verifier in a separate container scores the result.
The score
Each task scores between 0 and 1; the headline is the mean over tasks and five trials per task.
Test set
34 tasks (v2), as published by the authors.
Made by
Proximal, first published September 2026 (v1 in April 2026).
Official sources
- www.frontierswe.com
- Definition read on 2026-11-08
Sources and freshness
6 published scores from 4 sources. Each was read on its source page; the oldest reading on this page is from November 8, 2026. Prices beside each score come from our model price index and move when a vendor reprices.
- Vendor-reported Alibaba (Qwen)
- Vendor-reported Z.ai (Zhipu)
- Vendor-reported Google DeepMind
- Vendor-reported Moonshot AI
FrontierSWE FAQ
Answers derived from the 6 published scores on this page and the current price index.