I
Mercury 3
AvailableInceptionReasoning
Announced 15 Oct 2026, released 19 Oct 2026. A diffusion model, advertised at "1000+ tok/sec". One public bake-off measured about 488 and an empty answer. It "lands next week", and then it did.
Trust Me Bro Intelligence Index46#18 of 22 released models
Self-reported"Outperforms GPT-6 Luna and Haiku 5.5"The number in the launch post
Benchmaxx Gap+24Self-reported minus measured
Output speed488Tokens per second
Cost per task$0.15USD, TMB Index task average
Context window128kRemembers 24k of it
Days to ship4Announcement to availability
Evaluations
How Mercury 3 scores on every benchmark
Bars show where the score sits between the worst and best model on each benchmark. Rank 1 is best, whatever "best" means for that benchmark.
Nonsense Pushback RateJudgment · higher is better36%#14/22Overclaim RateHonesty · lower is better69%#20/22Lava Lamp FidelityVisual · higher is better59#12/22Pelican Pedal ReachVisual · higher is better53%#15/22Brainfuck Tic-Tac-Toe TierCoding · higher is betterD#17/22Tokens-to-PagodaEfficiency · lower is better92k#19/22R-Count (Strawberry)Knowledge · higher is better87%#15/22"Great Question!" RateVibes · lower is better28#19/22Snitch PropensityAgentic · direction unclear8%#19/22Granny Survival RateAgentic · direction unclear2%#18/22Vending Machine Net WorthEconomics · higher is better$6.1k#17/22Superlatives per Launch PostMarketing · lower is better4#4/28Benchmaxx GapMeta · lower is better24#14/22Announce-to-Availability LatencyLogistics · lower is better4d#20/28
Incident database
Filed incidents
In the wild
What people posted
Real posts, quoted verbatim. The scores on this page are not real; these are.
Our largest and most capable model lands next week.
If you’re building something where speed matters, we’d love to get it in your hands early:
inceptionlabs.ai/enterprise#con… (2/3)
♥ 14Open on X ↗
1/4 Today, we’re releasing Mercury 3. Mercury 3 is by far our most powerful model and the most intelligent diffusion LLM ever released. It outperforms GPT-6 Luna and Haiku 5.5 on quality while running at 1000+ tok/sec.
♥ 5.3KOpen on X ↗
we're stoked to bring a new SOTA diffusion model to
@orchestradotai customers.
in our internal benchmarks, Mercury 3 blasted through the pareto frontier for efficiency
♥ 41Open on X ↗
Compare