New to AdaL? Learn what AdaL CLI is → · Read the docs →

A lot landed in AdaL over the past week. Three new frontier reasoning models — and OpenAI's latest image model — are now available via /model. Three of them are 50% off for their first week. Claude Opus 4.7 (launched last week) is still live at half price too.

Kimi K2.6 — Opus-class reasoning at ~6× lower price

Moonshot's latest long-context model brings genuinely strong agentic and coding capability at a dramatic price reduction. In our internal testing, Kimi K2.6 approaches Claude Opus 4.6 on coding tasks at roughly 6× lower cost per token. Thinking is enabled, and the 224K context is enough for most real codebases.

Best for: teams that want Opus-class quality without Opus-class pricing. 50% off through May 2.

/model  # select kimi-k2.6

DeepSeek V4 Pro — strong reasoning, 936K context

DeepSeek V4 Pro is a deep reasoner with adaptive thinking and a very large context window. On output tokens, it's about 7× cheaper than Opus 4.6 — so long coding sessions with large generations get disproportionately cheap. Great when you need multi-file rewrites or long planning documents without flinching at the bill.

Best for: agentic coding loops with heavy output volume. 50% off through May 2.

DeepSeek V4 Flash — up to ~80× cheaper

V4 Flash is the budget end of the new DeepSeek lineup — designed for high-throughput agentic loops, data transformations, and subagent work. The output is up to ~80× cheaper than Opus 4.6 per million tokens. Thinking is enabled, and context stretches to 984K.

Best for: background jobs, subagents, bulk analysis. 50% off through May 2.

GPT Image 2 — design from inside your coding agent

OpenAI's latest image model is now routed automatically when AdaL needs to generate or edit an image. Photorealism, accurate in-image text, up to 4K output, every common aspect ratio.

The real unlock isn't the model — it's that your coding agent already knows your codebase, brand, and design system. Generating a landing-page hero or an OG card becomes "a hero illustration matching our design system" instead of a 300-word brief pasted into a separate tool. No tool switching, no re-pasting context — your agent does the whole asset pipeline.

Still live: Claude Opus 4.7 at 50% off

Anthropic's current flagship — 1M context native, adaptive thinking — is still at launch-week 50% off through April 27. If you haven't tried it yet, this weekend is a cheap window.

Try it

npm install -g @sylphai/adal-cli
adal

Then hit /model to pick any of the new additions. Full model list, pricing, and promotions:

→ All models & promotions in the docs

Share your feedback

Join the Discord community — we read everything, and the best model picks for your specific stack often surface there.

New Frontier Models in AdaL: DeepSeek V4, Kimi K2.6, and GPT Image 2

Subscribe to the Source

Comments