Mercury 3 [Inception Labs] - Matching Luna/Haiku at 1000 tokens/sec
The model supports a 1M‑token context window (up from 260K for Mercury 2.5) and is priced at $0.20 per million input and $0.70 per million output tokens, with a launch discount of 50% to $0.10 per million input and $0.35 per million output tokens; cached tokens are 90% off the input price.
Key points
- Launch price is $0.10 per million input and $0.35 per million output tokens, 50% off regular rates.
- Early partners report median generation 0.9 s, end‑to‑end 2.2 s, and 50% task completion on 50 Mind2Web tests.
Early‑access partners including TinyFish AI, Octen AI, Lorikeet and others report real‑time use cases. TinyFish AI uses Mercury 3 for voice tutoring with feedback returning in about a second. In a benchmark of 50 live Mind2Web navigation tasks, Mercury 3 completed 50% of the tasks, outperforming GPT‑6.1‑Sol while using the same token volume and agent steps. The model is available via the Inception API Platform, OpenRouter, Baseten and Vercel and is OpenAI‑API compatible.
Model pages: GPT-6 Luna → · Claude Haiku 5.5 → · GPT-6.1 Sol →
Mercury 3 [Inception Labs] - Matching Luna/Haiku at 1000 tokens/sec
inceptionlabs.ai · 19 October 2026
Loading the full article…
This text was published by inceptionlabs.ai. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
2sources- Reddit discussionreddit.com
- Inception Releases Mercury 3, Its Flagship Diffusion LLMPress · Unite.AI ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- DeepSeek’s Flash ties Anthropic’s Opus 5.5 in browser-agent benchmark · 2 src
- MiniMax releases M3.1 Flash model at $0.15 per 1M input tokens · 4 src
- Meta releases Muse Spark Open, an 816B-parameter open-weight multimodal model · 1 src
- Kling AI releases version 4.0 with stereo audio and a lighter Flash variant · 1 src
- DeepSeek releases V4.1-Flash-1017 update with lower cost and higher DeepSWE score · 26 src
Comments
via GitHub Discussions