Mercury 3 on BenchGecko · Inception's diffusion LLM with a 1M token context
Inception released Mercury 3 on 2026-10-19. List price is $0.20 input and $0.70 output per M tokens, 50% off at launch, with a 1M token context, up from 260,000 for Mercury 2.5.
What launched
Inception released Mercury 3 on 2026-10-19. Mercury 3 is a diffusion LLM: it generates and refines several tokens in parallel instead of writing one token at a time. Inception says it is more than 40 times the size of Mercury 1 and still runs at more than 1,000 tokens per second on NVIDIA GPUs, with 1,068 tokens per second quoted in the launch post.
It is available through the Inception API, Baseten, OpenRouter and Vercel, behind an OpenAI compatible API. The weights are not public. The model is listed on BenchGecko under Inception, and the row now carries the official launch date of 2026-10-19.
Specs
From Inception's documentation and the BenchGecko model row:
- Context · 1M tokens, stored as 1,000,000 on BenchGecko
- Max output · 131,072 tokens
- Input and output · text
- Reasoning effort · instant, low, medium and high
- Features · tool calling and structured outputs
- Weights · not released
Pricing on BenchGecko
Inception's list price for Mercury 3 is $0.20 per M input tokens and $0.70 per M output tokens. At launch the model is 50% off, at $0.10 input and $0.35 output. Cached input is 90% below the input rate, so $0.01 per M tokens during the promotion and $0.02 at list price. Inception has not said when the launch discount ends.
BenchGecko price fields follow the live rate on OpenRouter, so the model page shows the discounted $0.10 and $0.35 today. The Inception endpoint row on BenchGecko carries a 50% discount flag. When the promotion ends, the daily price sync picks up the new rate.
Earlier Mercury models on BenchGecko:
- Mercury 2.5 · released 2026-09-08 · 260,000 token context · $0.04 input and $0.15 output, with an 80% discount flag on the Inception endpoint, so $0.20 and $0.75 before discount
- Mercury 2 · released 2026-03-04 · 128,000 token context · $0.25 input and $0.75 output, no discount
Mercury 3 has 3.8 times the context of Mercury 2.5. Its list input price matches the undiscounted Mercury 2.5 rate and its list output price is $0.05 lower.
Compare rates · BenchGecko pricing.
Vendor-reported results
Inception says Mercury 3 is on par with or better than GPT-6 Luna and Claude Haiku 5.5 across a range of standard benchmarks while running 4 to 8 times faster. The launch post does not publish per benchmark scores. In one partner test quoted by Inception, TinyFish reports that Mercury 3 completed 50% of 50 live Mind2Web navigation tasks.
BenchGecko records these as vendor claims. No benchmark scores or Gecko Tests are attached to Mercury 3 yet. Mercury 2.5 carries 6 benchmark scores on BenchGecko and Mercury 2 carries 5.
Sources
Source: BenchGecko
Written automatically from BenchGecko's tracked data for this period; every number comes from the changelog, model, mindshare and pricing datasets. Checked against the source data before publishing.