The app
Models
The models of the chat app, in the private lane and in the open lane.
Every model in the picker of the app belongs to one lane, and the picker shows each one with its lane, its context window, and its price. The prices of every model are also on thinkr.sh/pricing.
Private lane
The private lane runs open-weight models inside trusted execution environments (TEEs) that your browser verifies before each chat. GPT-OSS 120B is the default, and the model of the free trial.
| Model | Maker | Context | About |
|---|---|---|---|
| GPT-OSS 120B | OpenAI | 131K | OpenAI's open-weight reasoning model, quick and light. |
| DeepSeek V4.1 Flash | DeepSeek | 524K | DeepSeek's fast open-weight model, with a long context. |
| Kimi K3 | Moonshot AI | 262K | Moonshot AI's largest open-weight model, for hard reasoning, code, and long tasks. |
| GLM-5.3 | Z.ai | 524K | Z.ai's open-weight flagship for coding and long, multi-step work. |
| Gemma 4 31B | 262K | Google's open-weight model for everyday tasks. | |
| Llama 3.3 70B | Meta | 131K | Meta's open-weight model for fast chat in many languages. |
Open lane
The open lane serves frontier and open-weight models through OpenRouter, under zero data retention (ZDR). Claude Haiku 5.5 and GPT-6 Luna are in the free trial.
| Model | Maker | Context | About |
|---|---|---|---|
| Claude Opus 5.5 | Anthropic | 1M | Anthropic's flagship model for hard reasoning, coding, and long, multi-step work. |
| Claude Sonnet 5.5 | Anthropic | 1M | Anthropic's balanced model for everyday writing, analysis, and code. |
| Claude Haiku 5.5 | Anthropic | 1M | Anthropic's small, fast model for quick answers and summaries. |
| GPT-6.1 Sol | OpenAI | 1M | OpenAI's upgrade to GPT-6 Sol for analysis, documents, and code. |
| GPT-6 Luna | OpenAI | 1M | OpenAI's fast, low-cost GPT-6 model for everyday chat. |
| Gemini 3.8 Flash | 1M | Google's fast Flash model with strong reasoning and coding. | |
| Grok 4.7 | SpaceXAI | 500K | SpaceXAI's flagship Grok model for coding and knowledge work. |
| DeepSeek V4.1 Flash | DeepSeek | 1M | DeepSeek's fast, low-cost model with a context of one million tokens. |
| GLM 5.3 Flash | Z.ai | 1M | Z.ai's fast multimodal model for coding and long tasks. |
| Kimi K3 | Moonshot AI | 1M | Moonshot AI's open-weight reasoning model for complex, long work. |
| Qwen3.8 2.4T | Qwen | 1M | Qwen's largest open-weight model, for hard reasoning and code. |
| Gemma 4 31B | 262K | Google's open-weight 31B model. |
DeepSeek V4.1 Flash, Kimi K3, and Gemma 4 31B appear twice in the picker, once in each lane. In the open lane they run outside a TEE, at the providers that OpenRouter picks, and their label says "Open · No retention".
Context
The context is the most tokens that one chat can hold in the app, as of 20 October 2026. Providers change their models over time.