Mercury 3

★★★★☆4.2 / 5 editorial Freemium Pay-as-you-go API: $0.20 per 1M input / $0.70 per 1M output tokens; 50% off at launch ($0.10 / $0.35); 100M free tokens for new accounts — check the official site for current pricing.

Diffusion LLM running at 1,068 tokens per second with frontier-class quality.

✓ Editor-researched listing — details reviewed by Grab AI Tool editors; pricing re-checked regularly.

Try it Free → ♡ 2,171 saves

#llm#api#diffusion#agents#fast

What is Mercury 3?

Mercury 3, launched October 19, 2026 by Inception Labs, is the most capable diffusion-based large language model released to date. Instead of generating text one token at a time like every mainstream model, diffusion lets Mercury write responses in parallel, and the company says the new model sustains 1,068 tokens per second on NVIDIA GPUs — roughly four to eight times faster than comparable speed-optimized models — with quality it claims matches or beats GPT-6 Luna and Claude Haiku 5.5 across its own benchmark suite.

Mercury 3 is over 40 times the size of Mercury 1, adds a 1-million-token context window (up from 260,000 in Mercury 2.5), and is available today through the Inception API, OpenRouter, Baseten, and Vercel. Because the API is OpenAI-compatible, developers can drop it into an existing stack without rewriting code. List pricing is $0.20 per million input tokens and $0.70 per million output tokens, with 50% off at launch and 100 million free tokens for new accounts — a direct shot at the frontier pricing war.

Who Mercury 3 Is Best For

Best for: Mercury 3 suits support, sales, and operations teams especially well: Extreme speed with claimed frontier-level quality breaks the usual speed-vs-intelligence trade-off; 1M context plus cheap cached tokens suits long agent loops and real-time apps. The free plan makes it easy to try on real work before paying anything.

Not ideal for: It may not suit you if any of these are dealbreakers: quality claims are Inception's own; one early review found it trailing Claude Haiku 5.5 on coding tasks; developer-only access today — there is no consumer chat app or playground.

Key Features

Diffusion architecture that generates tokens in parallel for 1,000+ tokens per second throughput
1-million-token context window for huge documents and long agent sessions
OpenAI-compatible API with model string mercury-3 and a reasoning_effort parameter
Live on Inception API, OpenRouter, Baseten, and Vercel with OpenAI-compatible endpoints
Benchmark suite (IFBench, GPQA Diamond, SWE-Bench Atlas) claimed on par with GPT-6 Luna and Claude Haiku 5.5

Pros

Extreme speed with claimed frontier-level quality breaks the usual speed-vs-intelligence trade-off
1M context plus cheap cached tokens suits long agent loops and real-time apps
OpenAI-compatible API makes switching from another model nearly effortless

Cons

Quality claims are Inception's own; one early review found it trailing Claude Haiku 5.5 on coding tasks
Developer-only access today — there is no consumer chat app or playground
Launch discount is time-limited; the standard rate doubles the per-token cost once it ends

Pricing

Model: Freemium
Pay-as-you-go API: $0.20 per 1M input / $0.70 per 1M output tokens; 50% off at launch ($0.10 / $0.35); 100M free tokens for new accounts — check the official site for current pricing.

Adding Up Your AI Bill?

If you pay for Mercury 3 alongside other AI subscriptions, the monthly total adds up quietly. Tick every tool you pay for in our free AI Tool Cost Calculator to see your real monthly and yearly bill — and spot overlaps you can cut. Still choosing? Start with our guide on how to choose the right AI tool.