Models & Billing
AdaL gives you access to the best AI models from leading providers. Switch models instantly with /model to match your task needs and budget.
See Pricing & Usage for included plan credits, weekly limits, usage tracking, top-ups, and bringing your own model access.
Switching Models
/model
Use /model to browse four sections:
- Recommended: curated top picks for most workflows
- New: recently added models
- Providers: full model lists grouped by provider
- Third Party Subscriptions: OAuth-based third-party model access, such as ChatGPT Subscription
Your selection persists across future AdaL sessions in this project.
Model Promotions
Active limited-time discounts. Pricing adjustments are applied automatically — no promo code needed.
| Model | Discount | Window | Slug |
|---|---|---|---|
| GPT-6 Astra | 50% off (launch week) | 2026-09-05 → 2026-09-12 | openai-gpt-6-astra |
| GPT-5.6 Terra | 20% off (through Dec 31, 2026) | 2026-07-30 → 2026-12-31 | openai-gpt-5.6-terra |
| GPT-5.6 Luna | 80% off (through Dec 31, 2026) | 2026-07-30 → 2026-12-31 | openai-gpt-5.6-luna |
| GPT-5.6 Sol | 20% off (through Dec 31, 2026) | 2026-08-21 → 2026-12-31 | openai-gpt-5.6-sol |
| Gemini 3.8 Flash | 50% off (Introductory pricing through Dec 31, 2026) | 2026-09-02 → 2026-12-31 | google-gemini-3.8-flash |
| Gemini 3.7 Flash | 50% off (Introductory pricing through Dec 31, 2026) | 2026-08-13 → 2026-12-31 | google-gemini-3.7-flash |
Promotions are sourced from the model registry and may update without notice. Check back here for the latest.
Recommended Models
These are our top picks, balancing capability, speed, and cost:
| Model | Provider | Context | Best For |
|---|---|---|---|
| Claude Opus 4.7 🆕 | Anthropic | 1M | Most capable: complex reasoning, agentic coding. 50% off launch week (ends Apr 27) |
| GPT-5.3 Codex | OpenAI | 272K | Coding-optimized for long-horizon tasks (default, price baseline) |
| Claude Sonnet 4.6 | Anthropic | 200K | Daily coding (slightly more expensive) |
| Claude Opus 4.6 | Anthropic | 200K | Complex reasoning, production code (2x more expensive) |
| Gemini 3.1 Pro | 1M | Multi-modal reasoning, design tasks (slightly cheaper) | |
| Gemini 3 Flash | 1M | Ultra-fast, simple tasks (4x in / 5x out cheaper) | |
| GLM-5 | Zai | 200K | General coding and reasoning (2x in / 5x out cheaper) |
| GLM-4.7 FlashX | Zai | 200K | Fast budget coding (29x in / 40x out cheaper) |
All Models by Provider
Legend: ⭐ recommended · 🆕 new
Anthropic (5 models)
- Claude Fable 5.1 — Demanding reasoning & long-horizon agentic work Slug:
anthropic-claude-fable-5-1 - ⭐ Claude Sonnet 5 — Most capable Sonnet · coding & agents Slug:
anthropic-claude-sonnet-5 - Claude Sonnet 4.6 — Proven all-rounder · coding & agents Slug:
anthropic-claude-sonnet-4-6 - ⭐ Claude Opus 5 — Most capable Anthropic · complex agentic coding Slug:
anthropic-claude-opus-5 - ⭐ Claude Opus 4.6 — Deep reasoning & production code Slug:
anthropic-claude-opus-4-6
OpenAI (4 models)
- GPT-6 Astra — Frontier · hardest end-to-end reasoning and coding work 50% off launch week through 2026-09-12. Slug:
openai-gpt-6-astra - ⭐ GPT-5.6 Terra — Balances intelligence and cost 20% off through Dec 31, 2026 through 2026-12-31. Slug:
openai-gpt-5.6-terra - ⭐ GPT-5.6 Luna — Cost-sensitive high-volume workloads 80% off through Dec 31, 2026 through 2026-12-31. Slug:
openai-gpt-5.6-luna - ⭐ GPT-5.6 Sol — Frontier · complex professional work 20% off through Dec 31, 2026 through 2026-12-31. Slug:
openai-gpt-5.6-sol
Google (5 models)
- Gemini 3.8 Flash — Most capable Flash for coding/agentic workflows · 1M context Introductory pricing through Dec 31, 2026 through 2026-12-31. Slug:
google-gemini-3.8-flash - ⭐ Gemini 3.7 Flash — Most capable Flash for coding/agentic workflows · 1M context Introductory pricing through Dec 31, 2026 through 2026-12-31. Slug:
google-gemini-3.7-flash - Gemini 3.1 Pro — Best multimodal understanding · 1M context Slug:
google-gemini-3.1-pro-preview - Gemini 3.6 Flash — Fast multimodal · large output budget Slug:
google-gemini-3.6-flash - Gemini 3 Flash — Fast everyday multimodal · 1M context Slug:
google-gemini-3-flash-preview
Z.ai GLM (4 models)
- ⭐ GLM-5.3 Flash — Native multimodal · flash-cost coding Slug:
zai-glm-5.3-flash - ⭐ GLM-5.3 — Flagship · long-horizon coding Slug:
zai-glm-5.3 - GLM-5.2 — Previous-gen GLM flagship · coding Slug:
zai-glm-5.2 - GLM-5.1 — Previous-gen GLM flagship · coding Slug:
zai-glm-5.1
MiniMax (2 models)
- MiniMax M2.7 — Lightweight agentic coding Slug:
minimax-MiniMax-M2.7 - ⭐ MiniMax M3 — Good with browser-use · coding & agentic Slug:
minimax-MiniMax-M3
DeepSeek (3 models)
- DeepSeek V4 Flash — Fast reasoning · 1M context Slug:
deepseek-deepseek-v4-flash - ⭐ DeepSeek V4 Flash Vision — Experimental · vision-capable DeepSeek V4 Flash Slug:
deepseek-deepseek-v4-flash-vision-exp - DeepSeek V4 Pro — Previous flagship · frontier reasoning · 1M context Slug:
deepseek-deepseek-v4-pro
Moonshot (1 models)
- ⭐ Kimi K3 — Frontier coding · deep reasoning · 1M context Slug:
kimi-kimi-k3
ChatGPT Subscription (OAuth) (3 models)
- GPT-5.6 Sol — Frontier · complex professional work Slug:
chatgpt_web-gpt-5.6-sol - GPT-5.6 Terra — Balances intelligence and cost Slug:
chatgpt_web-gpt-5.6-terra - GPT-5.6 Luna — Cost-sensitive high-volume workloads Slug:
chatgpt_web-gpt-5.6-luna
Image Models
AdaL also supports image generation and editing. Ask AdaL to generate or edit an image and it will route to the appropriate model automatically.
These are tool models, not planner models — they don't appear in /model. The slug below is the model name the image tool accepts, so you can request a specific one ("generate this with Nano Banana Pro").
- ⭐ Nano Banana 2 (Google) — Best all-around image generation, instruction following Slug:
nano-banana-2 - ⭐ Nano Banana Pro (Google) — Professional assets, text rendering, up to 4K Slug:
nano-banana-pro - ⭐ GPT Image 2 (Openai) — OpenAI image generation with high fidelity and text rendering Slug:
gpt-image-2
| Model | Max input images | Resolutions |
|---|---|---|
| Nano Banana 2 | 3 | 512px, 1K, 2K, 4K |
| Nano Banana Pro | 14 | 1K, 2K, 4K |
| GPT Image 2 | 1 | 1K, 2K, 4K |
Video Models
AdaL supports AI video generation from Google's Veo (the default) and MiniMax's Hailuo 03. Ask AdaL to generate a video and it handles resolution, duration, and polling automatically. Name a model to steer it — for example, "use minimax-h3 for a 12-second clip".
Like the image models above, these are tool models — not selectable via /model.
- ⭐ Veo 3.1 (Google) — Cinematic video generation with native audio Slug:
veo-3.1 - MiniMax H3 (Minimax) — Hailuo 03: up to 15s clips at 2K with native stereo audio Slug:
minimax-h3
| Model | Slug | Resolutions | Duration | Aspect ratios | Cost |
|---|---|---|---|---|---|
| Veo 3.1 (default) | veo-3.1 | 720p, 1080p, 4K | 4, 6, or 8 seconds (4K: 8 only) | 16:9, 9:16 | ~$0.40/sec (4K ~$0.60/sec) |
| MiniMax H3 | minimax-h3 | 768P, 2K | any whole number from 4 to 15 seconds | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 | $0.08/sec at 768P, $0.13/sec at 2K |
Capabilities: both models do text-to-video, image-to-video, frame interpolation (start + end frame), and reference images for subject/style consistency (Veo up to 3, MiniMax H3 up to 9). Video extension is Veo-only and renders at 720p. Audio is generated natively from the prompt on both. MiniMax H3's aspect ratio applies to text-to-video; image-to-video follows the input image.
→ Full capabilities guide: Image & Video Generation
Third Party Subscriptions
ChatGPT Subscription (OAuth)
Use your existing ChatGPT Plus/Pro subscription to access Codex models in AdaL — no API key needed. You’ll see this under /model → Third Party Subscriptions.
→ Setup guide: ChatGPT Subscription
Local Models (Preview)
Run AI models entirely on your machine — no API key, no cloud costs, no data leaving your device.
AdaL supports local models via Ollama. Once Ollama is running with a model pulled, select it from /model under the Local section.
/model # scroll to Ollama section → select a model
Supported models include GPT-OSS 20B and Qwen3-Coder 30B. Local models are free to use but require a capable GPU/CPU.
→ Full setup guide: Local Models with Ollama
Key Features
Adaptive Thinking & Effort Control
All models use adaptive thinking that automatically scales reasoning depth based on task complexity. Thinking is always on by default and adjusts itself — perfect for debugging, architecture decisions, and complex refactoring.
You can also manually tune the thinking effort level with /model config:
/model config
This opens an interactive dialog right in your terminal. Use arrow keys (↑↓) to browse the available effort levels for your current model, then press Enter to confirm. The dialog shows a description for each level so you know what you're picking:
| Level | Behavior |
|---|---|
| max | Always thinks with no constraints on depth |
| high | Always thinks deeply (default) |
| medium | Moderate thinking — may skip for simple queries |
| low | Minimal thinking — fastest, lowest cost |
The available levels depend on the model — not all models support every level. The dialog only shows what's valid for your current selection.
Shortcut: You can also skip the dialog and set it inline: /model config effort=high.
Your effort setting persists per project. Lower effort = faster responses and lower token cost for simple tasks.
Prompt Caching
Reusing context (files, conversation history) costs 50-90% less with cached inputs. Caching is automatic — AdaL handles it behind the scenes.
Extended Context
Handle large codebases with models supporting up to 1M tokens:
- Claude Sonnet 4.6 (1M) / Opus 4.6 (1M)
- Gemini 3.1 Pro / 3 Pro / Flash / 2.5 Pro
Perfect for reviewing entire repositories or understanding complex systems.
Billing
AdaL offers two billing options:
-
AdaL CLI Subscription — Subscribe with monthly credits included. Use any model seamlessly—credits are deducted automatically based on token usage.
-
Pro + BYOAK (Bring Your Own API Key) — Use your own API keys for supported providers while maintaining a Pro subscription (or higher) to ensure all features work seamlessly.
See Pricing for subscription tiers and credit details.
Pricing Reference
All models use pay-per-token pricing based on input and output tokens. Prompt caching reduces costs by 50–90% on repeated context.
For official pricing from each provider:
Related: Quickstart · Input Methods · BYOAK · ChatGPT Subscription