Sunlight 2
Aug 20, 2026
Sunlight 2 released
Our fastest frontier model — 3× cheaper than Sunlight 2 Pro, with strong SWE-bench and HumanEval scores. Ideal for autocomplete, search and fast agents.
Fast and affordable
Sunlight 2
A fast frontier model for everyday coding, search and interactive edits.
Balanced speed and quality — 3× cheaper, 200K context.
Sunlight 2Try Sunlight 2
Starting at
$1.20 per 1M input tokens
Announcements
Model card
Aug 20, 2026
Sunlight 2 Model Card
Official model card for Sunlight 2. Documents training data, evaluation methodology, benchmarks and safety results.
Meteor × Open Compute
Jul 14, 2026
Meteor × Open Compute — faster inference
We partnered to accelerate training and inference, bringing Sunlight 2 to all Meteor users with < 300ms median latency.
Use cases
Autocomplete
Sub-300ms completions with deep repo context. 40% acceptance rate.
Code search
Natural language search across your codebase with semantic ranking.
Fast agents
Cheap enough to run on every keystroke, smart enough to plan edits.
Overview
Sunlight 2 is Meteor's workhorse — optimized for speed without sacrificing frontier quality. Trained on 12T tokens of code and text, with a 200K context window and aggressive distillation, it hits 92.1% on HumanEval and 62.8% on SWE-bench Verified while being 3× cheaper than our Pro model. Ideal for interactive use: Tab autocomplete, Cmd-K edits, and background agents that need to be fast and precise.
200K
context
3×
cheaper than Pro
<300ms
median latency
Foundation
Built on the same architecture as Sunlight 2 Pro but distilled and fine-tuned for latency. Uses grouped-query attention, 8k native generation and tool-use pre-training. Trained with RL on real Meteor trajectories: 400k agent episodes with test-driven rewards. Fully evaluated for safety with no regression on our harm benchmarks.
Benchmarks
Sunlight 2 vs best competitor at similar price. Higher is better.
SWE-bench Verified
62.8%
54.2%
HumanEval
92.1%
90.2%
MMLU
86.7%
86.1%
GPQA Diamond
52.3%
49.8%
AIME 2024
38.5%
30.1%
LiveCodeBench
61.4%
55.9%
Sunlight 2 Best competitor
| Benchmark | Sunlight 2 | GPT-4o | Claude 3.5 Sonnet |
|---|---|---|---|
| SWE-bench Verified | 62.8% | 48.1% | 54.2% |
| HumanEval (pass@1) | 92.1% | 90.2% | 89.7% |
| MMLU | 86.7% | 86.1% | 85.4% |
| LiveCodeBench | 61.4% | 53.2% | 55.9% |
| Cost / 1M tokens | $1.20 | $3.50 | $3.00 |
Surfaces
Meteor IDE
Tab, Cmd-K, Agent and inline edits. Default for all users.
API
api.meteor.sh/v1 • OpenAI-compatible. Streaming + tools.
CLI
meteor agent --model sunlight-2 for terminal agents.
Frequently asked questions
When should I use Sunlight 2 vs Pro?
Use Sunlight 2 for interactive work and cost-sensitive agents. Switch to Pro for long-running tasks, vision, and hardest repos where quality matters over speed.
Is Sunlight 2 available via API?
Yes — same API as Pro, just set model: 'sunlight-2'. Pricing is $1.20 / $4.80 per 1M input/output tokens.
Does it support vision?
Limited vision support. For heavy visual work, use Sunlight 2 Pro which has frontier vision.
Start building with Meteor
Download for Mac or Windows. Start free in 30 seconds.