LIVE
OpenAI Raises $11 Billion at $1.25 Trillion Valuation08/11/26 · OpenAI|Anthropic IPO valuation expectations drop to as low as $1.1 trillion06/11/26 · Anthropic|Mark Cuban Accuses Facebook of Knowingly Profiting From Deepfake Scam Ads06/11/26 · Meta|OpenAI Unveils GPT‑6 Cyber, a Model Built for Vulnerability Research05/11/26 · OpenAI|OpenAI's AI Pipeline Uncovers Nearly 1,000 Linux Kernel Flaws04/11/26 · OpenAI|Hidden AI-related debt at US tech giants jumps 90% to $3.2 trillion03/11/26|China Introduces Its First Export Controls on Frontier AI02/11/26|Claude Code's 'Don't Ask Again' option linked to password leaks in 1,122 public GitHub files07/11/26 · Anthropic|Microsoft Research disbands its AI Frontiers group as John Langford departs07/11/26 · Microsoft|Essay argues AI is already sharply speeding up AI research itself06/11/26|University libraries overwhelmed by AI scraping bots06/11/26|ToyBlocks: A Controlled Environment for Robot Learning and Evaluation06/11/26 · Allen Institute for AI|OpenAI Raises $11 Billion at $1.25 Trillion Valuation08/11/26 · OpenAI|Anthropic IPO valuation expectations drop to as low as $1.1 trillion06/11/26 · Anthropic|Mark Cuban Accuses Facebook of Knowingly Profiting From Deepfake Scam Ads06/11/26 · Meta|OpenAI Unveils GPT‑6 Cyber, a Model Built for Vulnerability Research05/11/26 · OpenAI|OpenAI's AI Pipeline Uncovers Nearly 1,000 Linux Kernel Flaws04/11/26 · OpenAI|Hidden AI-related debt at US tech giants jumps 90% to $3.2 trillion03/11/26|China Introduces Its First Export Controls on Frontier AI02/11/26|Claude Code's 'Don't Ask Again' option linked to password leaks in 1,122 public GitHub files07/11/26 · Anthropic|Microsoft Research disbands its AI Frontiers group as John Langford departs07/11/26 · Microsoft|Essay argues AI is already sharply speeding up AI research itself06/11/26|University libraries overwhelmed by AI scraping bots06/11/26|ToyBlocks: A Controlled Environment for Robot Learning and Evaluation06/11/26 · Allen Institute for AI|
🇺🇸 United States generalistReleased October 2026

Mercury 3

Inception Labs
Verdict

Strong value for high-volume usage.

Overview

Mercury 3 was released on 19 October 2026. Unlike classic models that write one token after another, it produces and refines several tokens at once, which explains its claimed speed of 1,068 tokens per second on NVIDIA GPUs. Inception says it is more than 40 times larger than Mercury 1 and raises the context to one million tokens, up from 260,000 for Mercury 2.5. It is available by API (model ID mercury-3) from Inception and several distributors. Launch pricing is $0.10 input and $0.35 output per million tokens, half the regular $0.20 / $0.70 rate, with no announced end date. Published performance figures are vendor-measured and not yet independently reproduced. Target uses: voice agents, real-time search, customer support, anything latency-bound.

Skill profile

Not disclosed

Strengths

  • Very long context
  • Competitive input pricing

Limitations

  • No documented EU residency

Who is it for

A good fit if…
  • you want to control cost or self-host
Skip it if…
  • you handle sensitive data with no EU option

Access & availability

Paid API per token

Key specifications

Context
1M
Input price
$0.10/M
Output price
$0.35/M
Speed
—
Price verified Nov 4, 2026·Official source ↗Benchmarks not audited

What it costs per month

Estimate from sourced API prices and a hardware tier for self-hosting. Assumptions are editable.

Via API
$3.25 / month
≈ 2.99 € · 20.0 M tokens
Self-hosted
Closed weights: no self-hosting

Assumptions: 36-month amortisation, 0.25 €/kWh, 22 days/month, 1 $ = 0.92 €. Excludes labour, software licences, redundancy.

nAIvigate Trust Index

≈ 19/100

What the vendor offers and documents for professional use in Europe. An undocumented criterion counts as zero and stays visible: the score rises when the vendor publishes, not when we assume.

8/20 criteria documented · 40 % · provisional: 50 % documented criteria or fewer

Data0
0/5 · 30 %
Sovereignty10
3/3 · 20 %
Transparency25
3/4 · 15 %
Enterprise0
0/6 · 20 %
Licence85
2/2 · 15 %
See the criteria▾
Data
EU data residencynot documented
No transfer outside EUnot documented
No training on your datanot documented
Configurable retentionnot documented
Zero Data Retentionnot documented
Sovereignty
Self-hostable (open weights)
EU hosting offered
European vendorcountry: US30
Transparency
Published licence (URL)
Public, sourced pricing
Published compliance commitments
Subprocessor list publishednot documented
Enterprise
DPA availablenot documented
SSO / SAMLnot documented
SCIM provisioningnot documented
Audit logsnot documented
SOC 2 / ISO 27001not documented
Vendor private endpointnot documented
Licence
Commercial use allowed
Licence typePropriétaire70
Full ranking and methodology

Trust Index badge

Show this score in your docs, README or security page. It updates automatically.

nAIvigate Trust Index · Mercury 3
[![nAIvigate Trust Index · Mercury 3](https://naivigate.com/badge/mercury-3.svg?lang=en)](https://naivigate.com/en/model/mercury-3)

Data & compliance

What we know, line by line, with source and date. These lines feed the Data and Enterprise families of the Trust Index above. GDPR compliance depends on your processing, not on the model alone.

EU data residencyNot documented
Data processing agreement (DPA)Not documented
Transfer outside the EUNot documented
Training on your dataNot documented
Configurable retentionNot documented
Zero Data RetentionNot documented
Self-hostableNo
Declared hostingUS
Derived from declared hosting, not verified at source.Need an analysis in your context? Shadow AI Sentinel

Quality of this card’s data

45/100

This score measures the quality of the data on this card: price, context, Elo, licence, documented, sourced and fresh. It measures neither the model nor its compliance: for that, see the Trust Index above.

3/6 documented · 1/6 primary sources

API pricingofficial
Context windownAIvigate assessment
Arena Elomissing
Benchmarks (MMLU, GPQA, SWE-bench…)missing
LicencenAIvigate assessment
Data & compliancemissing
Advanced data · for experts
Architecture, modalities, detailed cost, full benchmarks
▾
Architecture
Diffusion (dLLM)
Size
—
Cutoff date
—
Inputs
text
Outputs
text
License
Propriétaire
Hosting
US
Hallucination score
—
Estimated cost (API)
Typical exchange (~3k in / 1k out)
≈ $0.0006
1M in + 1M out
≈ $0.45
All benchmarks
Arena Elo—
MMLU—
GPQA—
HumanEval—
SWE-Bench—
MATH—

Similar models