Skip to content

Inceptionlabs · Chat

Mercury 2

InceptionlabsChatMercury familyOne of its sources says the weights aren't publicly available.

Newer version: Mercury 3. We compared release dates and version names in our catalogue.

From Inception's listing:
Mercury is the first diffusion large language model (dLLM). Applying a breakthrough discrete diffusion approach, the model runs 5-10x faster than even speed optimized models like Claude 3.5 Haiku and GPT-4o Mini while matching their performance. Mercury's speed means that users can stay in the flow, enjoying rapid chat-based iteration and responsive completion suggestions. On Copilot Arena, Mercury ranks 1st in speed and ties for 2nd in quality.
llminfo IQ
116.5
Market median price
$0.375 blended / 1M
Max context
128K
Listings
4

Evidence and sources

llminfo IQ 116.5, ranked among chat models. From 53 benchmarks. Plenty of data behind it.How llminfo IQ is computedCoding IQ 93.2, 95% interval 80–106. From 6 benchmark families across 4 sources. A fair amount of data behind it. Uses the model's best published results. It has its own scale, separate from llminfo IQ.How Coding IQ is computedGood value (56.7). Cheaper per task than its benchmark results predict. From 4 benchmarks with measured task costs. A fair amount of data behind it.How Value for money worksYap score: tokens used per coding task vs. models with the same resultsTerse (43 ±9). Generates fewer tokens than models with similar coding results. Measured at default effort. Little data behind it, so treat it as a rough estimate.Median 9,229 output tokens per task at default effort.How the Yap score worksReleased 24 Feb 2026
All rankings: #401 on OpenRouter this month #589 on NanoGPT recentlySources: Inception Labs data, Models.dev data, llminfo researchPrimary data from Inception Labs from Models.dev: weights availability from llminfo research: release date

Advantages and trade-offs

Compared with 10 chat models, picked mainly by popularity and IQ

Trade-offs

No open weights

The weights aren't public, so you can't run it yourself.

We only make a claim when there's enough data on models of the same type. Updated 22 Nov, 06:07 UTC.Prices come from the model's maker where they sell it, otherwise their official partner, otherwise the market median. This comparison uses 9 maker's prices; 1 market median price.

Input & output

as stated by the model's providers
Text
Text

Capabilities

Supported operations and provider-stated features

ReasoningFunction callingStructured outputPrompt caching

Listings

 Flowdotinception/mercury-2Source: FlowDot data128K$0.25$0.75
 NanoGPTmercury-2Sources: Models.dev data, NanoGPT128K$0.25cache read $0.025$0.75
 OpenRouterinception/mercury-2Source: GenAI Prices dataNot published by the provider$0.25cache read $0.025$0.75
 InceptionRegions: 🇺🇸 USmercury-2Sources: Inception Labs data, LiteLLM data, Models.dev data128K$0.25cache read $0.025$0.75

Benchmarks

Named effort levels run from lowest to highest. Select a result to see its source details.

Human preference

Arena ELO24 related benchmarks · 24 results
Design Arena EloSource: Design Arena Leaderboards data · Nov 2026

Coding

Arena ELOSource: LMArena data · Nov 2026

Agentic

FRAMES · Agent ScoreMercury 2 official search evaluation · n=100 · medium reasoning · Exa loop · 25 calls · GPT-5.4 judge · Reported by Inception Labs · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.

Factuality

Instruction following

IFBenchSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim

Science

GPQASource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim

General

Math

AIME 2025Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim

Reasoning

Other

DeepSearchQA · F1 RatioMercury 2 official search evaluation · n=100 · medium reasoning · Exa loop · 30 calls · Gemini 2.5 Flash judge · Reported by Inception Labs · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
PinchBench3 related benchmarks · 3 results
v2 · Best Success Rate
Reported by PinchBench · via BenchBoard docs
OpenClaw · PinchBench v2.0.0 · task count: 147 on live leaderboard / 148 in v2.0.0 release notes · best run
v2 · Average Success Rate
Reported by PinchBench · via BenchBoard docs
OpenClaw · PinchBench v2.0.0 · task count: 147 on live leaderboard / 148 in v2.0.0 release notes · average across 20 runs
Version/Metric Unspecified
Reported by Inception Labs · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Mercury 2 official PinchBench evaluation · OpenClaw · success rate · version not stated

Data credits: Data by LLM Stats · Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · LMArena, CC BY 4.0 (licence) · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · SnakeBench (Greg Kamradt / ARC Prize Foundation), MIT (licence) · Vectara Hallucination Leaderboard, Apache-2.0 (licence)

Also known as: inception/mercury-2

Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works