Short answer
Gestalt is a stealth model: nobody has said who made it, and you can't use it yet. On November 19, 2026, Contra Labs launched the Frontier Design Benchmark, where professional designers vote blind on websites built from real client briefs. Gestalt came first, ahead of every named model and of the human designers' own reference designs. Contra calls it "a high-quality design model currently in stealth" and hasn't named the lab behind it.
- What's confirmed: Gestalt's scores on Contra's benchmark, from Contra's launch post and live leaderboard. Its maker, model ID, price and release date are unknown, and it may never come as a public API.
- What you can use now: the models that rank right behind it on the same benchmark — Claude Fable 5.2 (claude-fable-5-2), Claude Opus 5.5 (claude-opus-5-5) and GPT-6.1 Sol (gpt-6.1-sol). All three work on TeamoRouter with one key. The Gestalt upcoming page tracks its status.
New accounts start with a $0 balance — top up before your first request. Top up from $5. If a request returns 402 (400 on the OpenAI format) with "wallet balance is insufficient", your balance is empty — your base URL and key are fine. Top up and retry.
Until Gestalt is out, use Claude Fable 5.2 or Opus 5.5 for design work with one key
What the Frontier Design Benchmark found
The benchmark has 33 design tasks: 27 based on real website jobs posted on Contra and 6 real projects from startups. Each model builds a working site for the brief, and professional designers recruited from Contra pick the better of two sites without knowing which model made which. Contra reports 4,311 votes from 35 designers so far.
From Contra's launch post:
- Gestalt scored 1,928 Elo and won 89% of its votes (378 of 423). The best model Contra ran itself, Claude Fable 5.2, scored 1,714. The two are 214 points apart and their 95% intervals don't overlap.
- It beat the human reference designs. When designers saw a Gestalt site next to a "golden" design made by a hired human designer, they picked Gestalt 20 times out of 28. It is the first model Contra has seen rated above the human designs.
- Its pages don't look AI-made. On Contra's "Slop Detector", which scores how AI-looking a page is from 0 to 100, Gestalt read 1.2 and the human designs 0.9. The three Claude models read 31 to 38, the two GPT models 41 to 53.
Contra says it is "working with the team to continuously evaluate its quality". Two caveats: the benchmark is Contra's own, and Gestalt's results haven't been checked by anyone else.
The leaderboard (as of November 20, 2026)
| # | Model | Maker | Elo | Win rate |
|---|---|---|---|---|
| 1 | Gestalt (stealth) | Unknown | 1,931 | 90% |
| 2 | Human designers | Recruited from Contra | 1,739 | 74% |
| 3 | Claude Fable 5.2 | Anthropic | 1,713 | 73% |
| 4 | Claude Opus 5.5 | Anthropic | 1,691 | 73% |
| 5 | GPT-6.1 Sol | OpenAI | 1,668 | 70% |
| 6 | Claude Sonnet 5.5 | Anthropic | 1,667 | 69% |
| 7 | Kimi K3.1 Pro | Moonshot AI | 1,663 | 71% |
| 8 | GPT-6 Astra | OpenAI | 1,638 | 66% |
Source: Frontier Design Benchmark leaderboard. The live numbers move slightly as votes come in; the launch post gave Gestalt 1,928 and Fable 5.2 1,714. Below the top eight there is a gap of about 200 points: Qwen3.8 Max, GLM-5.3 Prime, Gemini 4 Carbon, Grok 4.7 and DeepSeek V4 Pro all sit between 1,121 and 1,434.
Who made Gestalt?
Nobody knows yet. Contra hasn't named the lab, and no lab has claimed it. Posts on X guess at different companies, Anthropic among them (example), but none offers evidence. We won't guess.
"Gestalt" is a German word for a shape or whole form; in design it's the idea that people see a layout as a whole before its parts. The name fits a design model but says nothing about who built it.
What isn't known yet
Who made it, whether it's a general model or built only for design, whether it will ever be released, its model ID, price and context window, and whether it will come with an API at all. We'll update this article if a lab announces it.
Use these models today
For design and front-end work, the models right behind Gestalt on the same benchmark:
- Claude Fable 5.2 (claude-fable-5-2): #3 overall, the highest-ranked model you can call today. Official price $8 / $40 per million tokens. See Claude Fable 5.2 API.
- Claude Opus 5.5 (claude-opus-5-5): #4, close behind Fable 5.2 at half the list price, $4 / $20. See Claude Opus 5.5 API key and pricing.
- GPT-6.1 Sol (gpt-6.1-sol): #5, OpenAI's best entry, $2 / $10. See GPT-6.1 Sol API.
- Kimi K3.1 Pro (kimi-k3.1-pro): #7, in the same top group at $2 / $10. See Kimi K3.1 Pro API.
All four use the same TeamoRouter key and base URL (https://api.teamorouter.com/v1 for the OpenAI-compatible format), and work in coding tools such as Claude Code. Current prices are on the pricing page.
If Gestalt ships: if it comes to TeamoRouter after release, change only model in your request to the ID its maker publishes. The base URL and key stay the same.
We'll update this page if a lab announces it
If a lab announces Gestalt, we'll update this article with the maker, official name, model ID and price, and whether it's available on TeamoRouter. You can also vote for it on the Gestalt upcoming page.
Paying for usage
No subscription: you top up a balance from $5, and each request is charged by the tokens it uses, by card or with USDT, USDC or ETH. The same key works for Claude, GPT, Gemini and Kimi models.
FAQ
What is Gestalt?
An AI model shown in stealth on Contra Labs' Frontier Design Benchmark, where it ranks first for website design, ahead of Claude Fable 5.2 and the human reference designs. Contra describes it as a design model; nothing else about it is public.
Who made Gestalt?
Unknown. Contra hasn't said and no lab has claimed it. Guesses on X, including Anthropic, have no evidence behind them.
Can I use Gestalt now?
No. There is no public API, app or model ID. For design and front-end work, use Claude Fable 5.2 (claude-fable-5-2) or Claude Opus 5.5 (claude-opus-5-5), which rank next on the same benchmark.
When will Gestalt be released?
No date has been given, and it isn't confirmed that it will be released at all. We'll update this article if a lab announces it.
Related
- Gestalt upcoming page
- Claude Fable 5.2 API: model ID, price and what changed
- Claude Fable 5.2 vs Fable 5.1 vs Opus 5.5
- GPT-6.1 Sol API: price and how to call it
Models in this article
- GPT 6.1 SolPricing, latency, uptime →
- Claude Fable 5.2Pricing, latency, uptime →
- Claude Opus 5.5Pricing, latency, uptime →
Keep reading
- One API Key for DeepSeek, Gemini, Claude and GPT in Your Product: An OpenAI-Compatible LLM Router Setup (2026)2026-11-06
- How Much Does Claude Code Cost on an API Key? Real Session Costs by Model2026-11-05
- How to Pay for Claude: Payment Methods for Claude Pro and the API (Including Without a Card)2026-11-04