Inception releases diffusion LLM Mercury 3 at roughly 1,000 tokens per second
10/20/2026 — 10/21, 07:17·2 sources (all)·2 reports (all)
Key takeaways
- Inception has released Mercury 3, a diffusion-based LLM it says matches Luna and Haiku on quality while running 4-8x faster, generating roug…
Background & analysis
On October 20, 2026, Inception released Mercury 3, an LLM built on a diffusion-based architecture. According to Techmeme, citing a report from The Deep View, Inception claims that Mercury 3 performs comparably to the Luna and Haiku models while running 4x to 8x faster, generating roughly 1,000 tokens per second. The report also notes that the AI industry has bet heavily on LLMs and on the massive data centers needed to scale them. So far, the available information about Mercury 3 comes solely from Inception's own claims; the report does not disclose parameter counts, training data, a release schedule, or availability channels, and no third-party benchmark results have been provided.
AI-generated from 1 reports · updated 22 hours ago
Latest turnDiffusion language model startup Inception AI has released Mercury 3, its new flagship. In the company's own benchmarks it beat OpenAI's GPT-6 Luna and Anthropic's Claude Haiku 5.5 on eight of 12 tests while running four to eight times faster, SiliconANGLE reports.
Related AI tools
- Heat index
- 147
- All sources
- 2
- All reports
- 2
- First seen
- 23 hours ago
Reports on this story headlines open the original
Diffusion language model startup Inception AI has released Mercury 3, its new flagship. In the company's own benchmarks it beat OpenAI's GPT-6 Luna and Anthropic's Claude Haiku 5.5 on eight of 12 tests while running four to eight times faster, SiliconANGLE reports.
SiliconANGLEAI score 55
Inception has launched Mercury 3, a diffusion-based LLM that it claims is comparable to Luna and Haiku models while running 4x to 8x faster at roughly 1,000 tokens per second. The Deep View reported the launch.
TechmemeAI score 59
Other stories people are talking about
- Heat index 48Artificial Analysis launches HealthBench leaderboards1 sources
- Heat index 38Splitting rule files per package for Codex and Claude Code in a monorepo1 sources
- Heat index 36Meta: Claude Tool's High-Frequency Reads Led to Account Suspension1 sources
- Heat index 44Developer splits work across Claude, Grok, Codex and custom router1 sources
- Heat index 68Selection guide maps AI coding tools to specific workflows1 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion