Inception has released Mercury 3, a diffusion-based language model that the company claims outperforms GPT-6 Luna and Claude Haiku 5.5 while generating text four to eight times faster.
The startup describes Mercury 3 as the most capable diffusion language model released so far. Its speed relies on a different construction method. Standard models like GPT or Claude write text word by word, meaning each new token must wait for the previous one to finish. Diffusion models, usually associated with image generation, sketch a rough version of the entire answer first. They then refine this draft over several passes, allowing many words to be generated simultaneously.
Internal benchmarks show mixed results
Inception’s own data indicates Mercury 3 generates more than 1,068 tokens per second on Nvidia GPUs. By comparison, the company states Claude Haiku 5.5 manages about 240 tokens per second and GPT-6 Luna about 130. No independent tests have verified these figures yet.
The model leads rivals in eight of 12 tests according to Inception’s internal table. The gap is widest in instruction following. On IFBench, Mercury 3 scores 76.84 points, which is 15 more than GPT-6 Luna. It also leads clearly in autonomous coding on SWE-Bench Atlas with 51.08 points, while its rivals reach 33.33 and 32.26.
Claude Haiku 5.5 wins the longer coding test DeepSWE. It is also better at finding information in long texts. GPT-6 Luna edges ahead in two tests from Artificial Analysis. Inception’s short overview is more cautious, describing Mercury 3’s quality as on par or better.
According to Inception, Mercury 3 has more than 40 times as many parameters as Mercury 1 but runs just as fast. Performance has also tripled since the small Mercury 1 mini. The exact parameter count remains undisclosed.
Targets apps where small delays add up
Inception is pitching Mercury 3 mainly for AI agents, search, customer support, and tutoring. AI agents complete tasks in many separate steps and wait on the model at each one. Those small delays pile up fast.
Octen AI builds search features for smart glasses. Founder Kuan Zou says an answer that takes five seconds feels broken. He notes the search itself took only about 80 milliseconds and the language model was the bottleneck. With Mercury 3, the median response arrived after 0.9 seconds. The full process took 2.2 seconds.
Education company General Learning says it now delivers hints and corrections to students in about one second. Customer service provider Lorikeet reports faster response times along with better results. At TinyFish AI, co-founder Shuhao Zhang says Mercury 3 was given 50 navigation tasks on real websites. It solved 25, beating GPT-6.1-Sol. Zhang does not give a number for the OpenAI model.
Pricing may shift quickly
Regular pricing is $0.20 per million input tokens and $0.70 per million output tokens. Inception is cutting prices in half at launch. It has not said when the discount ends. After that, Mercury 3 would cost more for output than GPT-6 Luna and Claude Haiku 5.5, at least for shorter inputs, where both charge $0.50. For very long prompts, though, both rivals charge a lot more. Mercury 3 can handle up to 1 million tokens at once.
The model is available on Inception’s own platform as well as OpenRouter, Baseten, and Vercel. It uses the same API as OpenAI’s models, so Inception says developers can switch with little effort.
In February, Inception released Mercury 2, which it called the first diffusion model to reason through intermediate steps before answering. Before that, the startup raised $50 million. Backers include NVentures, M12, and Snowflake Ventures, the venture arms of Nvidia, Microsoft, and Snowflake.
Google is working on the same approach. Its Gemini Diffusion demo was followed in June by the open model DiffusionGemma. Google says it runs up to four times faster than the standard Gemma 4 but delivers slightly weaker results.
What it means
Developers building real-time applications face a new option for speed. The model allows for rapid iteration in systems like smart glasses or autonomous agents where latency is critical. Users can expect faster feedback loops, though the cost structure is currently discounted and will likely rise to match standard pricing tiers once the launch period ends.
