Mercury 2.5
The diffusion LLM that generates at 1,100+ tokens/sec
Details
- External ID
- 1245269
- Source
- PH
- Company
- —
- Product
- Mercury 2.5
- Website domain
- producthunt.com
- Launched
- Sept. 9, 2026
- Cohort
- —
- Upvotes
- 3
- Upvotes percentile
- 0.8905013897797733
- Tags
- API, Artificial Intelligence, Development
- Fetched at
- Sept. 10, 2026, 1:14 a.m.
- Updated at
- Sept. 10, 2026, 1:14 a.m.
Description
Mercury 2.5 is Inception's most capable diffusion language model yet, generating at over 1,100 tokens per second by producing text in parallel instead of one token at a time. It brings a 40% intelligence improvement over Mercury 2, 260K context, tunable reasoning, parallel tool calls, and structured JSON — built for latency-sensitive search, voice, coding, and agent workloads.
Enrichment
- Theme
- lightweight and on-device AI runtimes
- Vertical
- Horizontal
- Function
- Model & infra
- Audience
- Developer
- AI stance
- AI-native
- Project type
- Commercial product
- Normalized one-liner
- fast diffusion language model
- Manually corrected
- False
Could you build this?
No Developing a novel diffusion-based language model capable of 1,100+ tokens/sec requires cutting-edge machine learning research, massive GPU clusters, and custom model architecture design.
What it would actually take: Building a diffusion language model requires a distributed training pipeline using PyTorch, Megatron-LM/DeepSpeed, and a cluster of hundreds or thousands of H100 GPUs. The hard part is formulating a continuous or discrete diffusion formulation for parallel text decoding that matches autoregressive perplexity while scaling to 260K context. This demands world-class AI researchers and multi-million-dollar compute budgets.
Competitors
Other products that read as similar to this one — 33 launches clear the similarity bar, closest 8 shown.
Attention rank: #6 of 34 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 303 days after the earliest competitor.
- Mercury Edit 2 · ph · 2026-04-04 · 171 upvotes · similarity 0.50
- Tiny Diffusion · hn · 2025-11-10 · 172 upvotes · similarity 0.44
- "Be horse." · hn · 2026-04-30 · 10 upvotes · similarity 0.43
- REPI · github · 2026-09-28 · 9 upvotes · similarity 0.38
- Digital to Wave: Wave computing-Memory · ph · 2026-09-14 · 1 upvotes · similarity 0.37
- Lamb Labs: Custom Chips for AI Inference · yc · 2026-08-03 · 64 upvotes · similarity 0.35
- I built a tiny LLM to demystify how language models work · hn · 2026-04-06 · 915 upvotes · similarity 0.35
- tarski · github · 2026-09-28 · 22 upvotes · similarity 0.35
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a model & infra tool for Fintech yet.