Sunlight 2 Pro
Aug 20, 2026
Sunlight 2 Pro released
Frontier model for long-running agents. State-of-the-art on SWE-bench, AIME and vision tasks. Powers Meteor Agent with 1M context.
Frontier performance
Sunlight 2 Pro
A frontier model for long-running agents and ambitious interactive and visual work.
Our most capable model — 1M context, frontier vision and tool use.
Sunlight 2 ProTry Sunlight 2 Pro
Starting at
$8 per 1M input tokens
Announcements
Model card
Aug 20, 2026
Sunlight 2 Pro Model Card
Official model card documenting performance, benchmarks, safety evaluations and training details for Sunlight 2 Pro.
Meteor × SpaceX
Apr 21, 2026
Meteor + SpaceX — training frontier models together
We partner to accelerate model training for ambitious engineering work, from rockets to code.
Use cases
Long-running agents
1M context, 200+ tool calls, recovers from failures and finishes migrations.
Vision + code
Understands screenshots, Figma, and video — generates pixel-perfect UI.
Ambitious refactors
Plans 50+ files, runs tests, fixes failures iteratively.
Overview
Sunlight 2 Pro is our frontier — built for agents that run for hours and ship production code. With 1M tokens, native tool use, and state-of-the-art vision (86.2% on MMMU), it leads on SWE-bench Verified (71.4%) and AIME 2024 (52.1%). It reasons with verifiable traces and can self-correct over long horizons. If Sunlight 2 is the sprinter, Pro is the marathon runner.
1M
context
71.4%
SWE-bench Verified
86.2%
MMMU (vision)
Foundation
400B dense transformer with MoE inference, 32k native generation, and 16B active parameters. Trained on 18T tokens including 4T code + vision + trajectory data. RL with verifiable rewards on SWE-bench, terminal tasks and visual diffs. Architecture shares weights with Sunlight 2 but with 4× depth and extended reasoning. Safety-tuned with constitutional AI and red-team coverage.
Benchmarks
Frontier results — Sunlight 2 Pro vs strongest closed models. Higher is better.
SWE-bench Verified
71.4%
62.1%
HumanEval
94.3%
92.1%
MMLU
88.9%
87.2%
AIME 2024
52.1%
38.9%
GPQA Diamond
61.8%
53.4%
MMMU (vision)
68.9%
62.0%
Sunlight 2 Pro Best competitor
| Benchmark | Sunlight 2 Pro | Claude 4 Opus | GPT-4o | Gemini 2.0 |
|---|---|---|---|---|
| SWE-bench Verified | 71.4% | 62.1% | 48.1% | 55.3% |
| AIME 2024 (pass@1) | 52.1% | 38.9% | 30.1% | 34.5% |
| GPQA Diamond | 61.8% | 53.4% | 49.9% | 51.2% |
| MMMU (vision) | 68.9% | 62.0% | 58.4% | 65.1% |
| LiveCodeBench | 68.9% | 60.2% | 53.2% | 57.8% |
| MMLU | 88.9% | 87.2% | 86.1% | 87.5% |
Evaluations run with identical prompts, temperature 0, 1× attempt. SWE-bench Verified = 500 human-verified tasks. See model card for details.
Surfaces
Meteor Agent
Default for Plan/Build. Handles 50+ file refactors with 1M context.
API
api.meteor.sh/v1 • model: sunlight-2-pro • 1M input, 200K output
CLI & Cloud
meteor agent --model sunlight-2-pro + Meteor Cloud runners.
Frequently asked questions
How is Pro different from Sunlight 2?
Pro is 4× larger, with 1M context vs 200K, frontier vision (86.2% MMMU), and stronger reasoning (52.1% AIME). Use it when quality > cost.
Pricing?
$8 / $24 per 1M input/output tokens via API. Included in Meteor Pro and Enterprise with higher rate limits.
Does Pro replace Sunlight 2?
No — they share the same API. Sunlight 2 stays the default for Tab/autocomplete due to speed. Pro is for Agent/Plan.
Start building with Meteor
Download for Mac or Windows. Start free in 30 seconds.