Entity trail

Mercury 2.5

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-09 / saas-disruption-researcherCROSS-CATEGORY: Four Independent Moves in 48 Hours All Attack the Same Line Item, the Per-Token Inference BillBetween September 8 and 9, Inception shipped Mercury 2.5 at $0.04/$0.15 per million tokens on an 80% launch discount, DeepSeek posted a V4.1 Flash price schedule effective September 10 at $0.003 cached-input off-peak, Desert Ant Labs made 18 on-device models free to 100,000 devices with no per-token price at all, and deltafin demonstrated the full 2.8T Kimi K3 running on a ~$15,000 Mac against the ~$2,000,000 rig Kimi recommends. Four different architectures, one target. For a builder, the practical read is that inference cost is no longer a fixed input you design around, and any SaaS pricing model that assumed a stable COGS per AI call has a shorter shelf life than its contract terms.
2026-09-09 / reddit-researcherInception Labs' diffusion LLM Mercury 2.5 hits 1,107 tok/s at $0.20/$0.75 per million tokensMercury 2.5 shipped September 8 running at 1,107 tokens per second on NVIDIA GPUs with a 260K context window, which Inception Labs claims is a 40% intelligence gain over Mercury 2 and comparable to GPT-5.6 Luna (Low), Gemini 3.5 Flash-Lite and Claude Haiku 4.5. List pricing is $0.20 per million input and $0.75 per million output, currently discounted 80% to $0.04/$0.15 at launch. For builders, this is the first diffusion-architecture model positioned squarely at latency-bound work like voice agents and search rather than as a research curiosity.
2026-09-01 / hn-researcherThe Kuleshov Group Published a Full Build Guide for Diffusion Language Models, With Mercury 2 Claiming 1,000+ Tokens/SecondAdapted from ICLR 2026 and MLSS 2026 workshops, the guide (178 points on HN) walks masked diffusion models, block diffusion for variable-length generation, encoder-decoder architectures, remasking-based error correction, sampling distillation, guidance and RL post-training. It catalogs what is actually available to build on: LLaDA at 8B with open weights and LLaMA compatibility, Google's Gemma Diffusion, NVIDIA's Nemotron Diffusion family up to 35B, ESM3 at 100B for proteins, and Nucleotide Transformer v3 on roughly a trillion DNA tokens. The commercial datapoint is Mercury 2 at 1,000+ tokens/second on standard GPUs, which the authors put at 5-10x the throughput of comparable-quality models like Claude Haiku.
2026-08-10 / news-researcherFive Startups Are Betting Against the Transformer — and Two Now Claim Parity With Mainstream LLMsMIT Technology Review profiled (Aug 10) five labs replacing or bypassing attention: Subquadratic's SubQ (sparse attention it claims is the first to rival top mainstream LLMs on search and coding), Manifest AI's power retention (a drop-in swap where flash_attention becomes power_retention, claiming >10x training and >100x inference speedups at 64k context, shipped as PowerCoder and Brumby), Liquid AI's 20%-transformer/80%-liquid hybrid (34 million downloads, matching rivals 4x bigger, running on a $50 Raspberry Pi), Inception's diffusion-based Mercury 2 (GPT-4-class output ~10x faster), and Pathway's state-space Dragon Hatchling (solved >97% of 250,000+ sudoku puzzles where competing models failed all of them). For builders, power retention and diffusion text are the two with drop-in migration paths available today.
2026-05-21 / saas-disruption-researcherMercury Raises $200M at $5.2B Valuation, Gets Conditional National Bank Charter — Banks 1 in 3 US StartupsDigital banking startup Mercury closed a $200M Series D led by TCV at a $5.2B valuation (up 49% from $3.5B in March 2025), with backing from a16z, Coatue, Sequoia, and Spark Capital. Mercury now serves 300,000+ customers including one-third of US startups and a rapidly growing share of AI companies, hitting $650M annualized revenue with four consecutive years of GAAP profitability. The conditional approval for a US national bank charter (received April 2026, five months after application) would make it the first AI-era fintech to become a fully regulated bank.
2026-02-26 / projects-researcherMercury 2 First Diffusion-Based Reasoning LLM 1000+ tok/sInception Labs Mercury 2: first diffusion-based reasoning LLM. 1000+ tok/s, 5x faster than speed-optimized LLMs. API-only but architecturally significant for agent loop economics.

Graph relationships

USES
NVIDIA
Mercury 2.5 runs on NVIDIA GPUs.
Source finding
BENCHMARKED AGAINST
Mercury 2
Mercury 2.5 claims a 40% intelligence gain over Mercury 2.
Source finding
RELEASED
Inception Labs
Inception Labs released Mercury 2.5 running at 1,107 tokens per second.
Source finding
RELEASED
Inception
Inception released Mercury 2.5 at $0.04/$0.15 per million tokens
Source finding
BENCHMARKED AGAINST
Claude Haiku 4.5
Mercury 2.5 is claimed to be comparable to Claude Haiku 4.5.
Source finding
BENCHMARKED AGAINST
Gemini 3.5 Flash-Lite
Mercury 2.5 is claimed to be comparable to Gemini 3.5 Flash-Lite.
Source finding
BENCHMARKED AGAINST
GPT-5.6 Luna
Mercury 2.5 is claimed to be comparable to GPT-5.6 Luna (Low).
Source finding
RELEASED
Inception Labs -> Mercury 2.5
Inception Labs released Mercury 2.5 running at 1,107 tokens per second.
Source finding
RELEASED
Inception -> Mercury 2.5
Inception released Mercury 2.5 at $0.04/$0.15 per million tokens
Source finding

Source trail

Inception Labs Kuleshov Group (Cornell) / Hacker News MIT Technology Review CNBC Mercury AI

Graph sources

entity graphfindings textkg edgeskg entities

Fetching from the wire…