Best AI News today

Mercury 3 — AI News Today

LLM

Mercury 3 is a diffusion large language model from Inception, released on 20 October 2026. Instead of writing left to right one token at a time, it denoises whole blocks of text at once, which is how Inception gets 1,068 tokens per second with a 1M-token context window. On Inception's own table it leads GPT-6 Luna and Claude Haiku 5.5 on 8 of 12 benchmarks.

21 stories about Mercury 3

searching across sources...

About Mercury 3

Mercury 3 is a large language model. Mercury 3 is a diffusion large language model from Inception, released on 20 October 2026. Instead of writing left to right one token at a time, it denoises whole blocks of text at once, which is how Inception gets 1,068 tokens per second with a 1M-token context window. On Inception's own table it leads GPT-6 Luna and Claude Haiku 5.5 on 8 of 12 benchmarks. This page tracks 21 recent news stories about Mercury 3, curated from 30+ sources and updated every 15 minutes.

Mercury 3 — Coverage Momentum

0
This month
↓100%
vs last month
7
All-time
Oct 2026
Peak month
0
0
0
0
7
0
JunJulAugSepOctNov

First tracked Oct 2026. Most-citing sources: Medium (5), Marktechpost (1), The Low End Disruptor (1). Data aggregated by Best AI News Today from 30+ sources.

21 stories
1 of 3 Top Stories
Inception's Mercury 3 diffusion model claims to beat GPT-6 Luna and Claude Haiku 5.5 at a fraction of the wait
Top StoryMedia

Inception's Mercury 3 diffusion model claims to beat GPT-6 Luna and Claude Haiku 5.5 at a fraction of the wait

A scrappy startup is claiming to have built an AI model that outperforms OpenAI's latest GPT offering while running exponentially faster and cheaper, challenging the assumption that raw computational power determines artificial intelligence dominance. If Inception's Mercury 3 claims hold up under independent scrutiny, it could upend the emerging market for specialized AI applications and force the industry's giants to reconsider their strategy of scaling up at nearly any cost. The announcement comes as rivals are increasingly betting that efficiency, not just capability, will define the next generation of competitive advantage.

The Decoder faviconThe Decoder·
Related OpenAI AnthropicHaiku 5.5GPT-6 Luna Claude

11 sources

SiliconANGLE faviconSiliconANGLETechRepublic faviconTechRepublicMashable faviconMashable
Mercury 3: What 1,068 Tokens a Second Means for Real Work
Media

Mercury 3: What 1,068 Tokens a Second Means for Real Work

by Balthasar — Inception’s new diffusion language model makes a strong speed claim.

Medium faviconMedium·
Media

Inception’s Mercury 3: 40x Bigger, 8x Faster [2026]

shattered.io faviconshattered.io·
Media

Mercury 3: The diffusion model bringing zero-lag AI to life

YourStory.com faviconYourStory.com·
Mercury 3 Arrives. Meta Shrinks the Agent Router.
Media

Mercury 3 Arrives. Meta Shrinks the Agent Router.

by AI & Data Engineer — Five developments for teams choosing models, designing agents and planning compute, with the evidence each decision still needs.

Medium faviconMedium·
RelatedMetaMeta
10x faster generation, 1.4x faster workflow: a Python lab
Media

10x faster generation, 1.4x faster workflow: a Python lab

by SHIVAM DWIVEDI — Inception says Mercury 3 generates 1,068 tokens per second on NVIDIA GPUs [1]. For this lab, we’ll compare hypothetical rates of 100 and…

Medium faviconMedium·
Discussions

Mercury 3: Diffusion at scale

Article URL: https://www.inceptionlabs.ai/blog/introducing-mercury-3 Comments URL: https://news.ycombinator.com/item?id=50155245 Points: 258 # Comments: 92

Hacker News faviconHacker News·
Mercury 3 and Sailor Moon: Thinking about the mechanism of intelligence creation from 'Mercury'
Media

Mercury 3 and Sailor Moon: Thinking about the mechanism of intelligence creation from 'Mercury'

In the world of AI, I encountered a name that caught my attention.Mercury 3.It is a diffusion language model developed by Inception Labs.When I saw that name, 'Sailor Moon' suddenly popped into my ...

note faviconnote·
Mercury 3 launches on Venice, bringing parallel token generation to private AI
Media

Mercury 3 launches on Venice, bringing parallel token generation to private AI

Inception Labs' Mercury 3 diffusion model launches on Venice, claiming 1,068 tokens per second, a 1M-token context window and low launch pricing.

Crypto Briefing faviconCrypto Briefing·
Media

⚡️ Diffusion Is Coming for Language. And Mercury 3 Just Made It Real.

Seven months ago, I wrote about AI moving from scribes to revelations. Today, Inception’s Mercury 3 is turning that prediction into a… Continue reading on Medium »

Medium · Altan "Atabarezz" Atabarut faviconMedium · Altan "Atabarezz" Atabarut·
Inception says its Mercury 3 model beats GPT-6 Luna and runs far faster
Media

Inception says its Mercury 3 model beats GPT-6 Luna and runs far faster

Inception's Mercury 3 is "almost instant", its CEO says, at over 1,000 tokens a second. The diffusion model's benchmarks are self-reported.

The Next Web faviconThe Next Web·
RelatedGPT-6 Luna
What is Mercury 3? Is the AI That Doesn’t Write Left to Right Finally Ready for Real Work?
Media

What is Mercury 3? Is the AI That Doesn’t Write Left to Right Finally Ready for Real Work?

by Levent Bulusan — Inception’s diffusion model says it beats Claude Haiku 5.5

Medium faviconMedium·
Media

Inception Launches Mercury 3, Setting a New Pareto Frontier in Speed and Intelligence

Yahoo Finance faviconYahoo Finance·
This AI Doesn’t Write Left to Right: Mercury 3 and the Rise of Diffusion Language Models
Media

This AI Doesn’t Write Left to Right: Mercury 3 and the Rise of Diffusion Language Models

by Peeyush Kant Misra — A new model drafts its whole answer at once and then sharpens it. Here’s how that works, what the launch claims, and what the first…

Medium faviconMedium·
Is Diffusion LLM the speed demon as promised?
Media

Is Diffusion LLM the speed demon as promised?

by Zheng "Bruce" Li — We test Inception’s Mercury 3 against the fastest autoregressive models from Anthropic, Google Gemini, OpenAI, and Cerebras to find out

The Low End Disruptor faviconThe Low End Disruptor·
Media

Mercury 3 vs Mistral Large 4: Two Unfinished Models and the Promises Attached to Each

OrcaRouter faviconOrcaRouter·
Media

Inception Launches Mercury 3 With One-Million-Token Context Window

TOKENPOST faviconTOKENPOST·
Research

Inception Releases Mercury 3 Diffusion LLM at 1,068 Tokens per Second

Inception released Mercury 3, a diffusion LLM it says runs at 1,068 tokens per second with a 1M-token context. On Inception’s own table it leads GPT-6 Luna and Claude Haiku 5.5 on 8 of 12 benchmarks....

Marktechpost faviconMarktechpost·
Mercury 3 claims to outperform GPT-6 Luna, Claude Haiku 5.5 at 8 times the speed
Media

Mercury 3 claims to outperform GPT-6 Luna, Claude Haiku 5.5 at 8 times the speed

California-based AI research and product company Inception has unveiled Mercury 3, its latest diffusion ...

Interesting Engineering on MSN faviconInteresting Engineering on MSN·

Frequently Asked Questions

What is Mercury 3?

Mercury 3 is Mercury 3 is a diffusion large language model from Inception, released on 20 October 2026. Instead of writing left to right one token at a time, it denoises whole blocks of text at once, which is how Inception gets 1,068 tokens per second with a 1M-token context window. On Inception's own table it leads GPT-6 Luna and Claude Haiku 5.5 on 8 of 12 benchmarks.

What is the latest news about Mercury 3?

As of today, there are 21 recent stories about Mercury 3. Recent headlines include: Mercury 3: What 1,068 Tokens a Second Means for Real Work; Mercury 3 and Sailor Moon: Thinking about the mechanism of intelligence creation from 'Mercury'; This AI Doesn’t Write Left to Right: Mercury 3 and the Rise of Diffusion Language Models. Updated every 15 minutes.

What other LLM AI models are there?

Mercury 3 is part of the LLM model category. Browse all comparable AI llm models with live news mentions and benchmarks.