Search and RAG builders
Run repeated planning, rewriting, extraction, and summarization calls within one user query.
Independent tool overview
Mercury 2.5: features, pricing, and who it is for
Inception diffusion language model for low-latency reasoning and tool-using applications.
Visit the official Mercury 2.5 site ↗
- Type
- Diffusion language model
- Context
- 260K tokens
- Current promo
- $0.04 input / $0.15 output per 1M
- List rate
- $0.20 input / $0.75 output per 1M
Overview
What Mercury 2.5 is
Mercury 2.5 is Inception's production diffusion language model. The company positions it for workloads where a model may be called many times within one interaction: search planning, query rewrite, RAG, coding subagents, and voice pipelines. Its launch describes a 260K-token context window, tunable reasoning, parallel tool calls, and schema-aligned JSON output. Inception reports throughput of 1,107 tokens per second under its stated NVIDIA hardware conditions; that is a vendor benchmark, not a guarantee of application latency.
The official model page currently shows an 80% discount from $0.20/$0.75 to $0.04/$0.15 per million input/output tokens. The discount is promotional and could change. Mercury 2.5 is accessible via Inception API and named partners, with enterprise options for dedicated capacity. Its value lies in latency and cost across repeated calls, but teams should test their own prompt quality, tool reliability, and total pipeline time.
Use cases
Who Mercury 2.5 is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Voice-agent teams
Place fast text reasoning behind a speech interface where audible pauses matter.
Coding-platform teams
Use a fast model for context compaction, tool search, routing, or helper agents.
Capabilities
Core Mercury 2.5 features
Diffusion model
Inception uses a diffusion language-model architecture designed for high-throughput generation.
Tunable reasoning
Developers can trade some speed for more reasoning effort on harder calls.
Parallel tool calls
The release lists concurrent tool use for agent workflows.
Structured JSON
Schema-aligned output supports extraction and application integration.
Large context
The vendor lists a 260K-token context window.
Process
How the Mercury 2.5 workflow works
Step 1
Identify repeated calls
Map the model steps in the existing workflow and isolate latency-sensitive ones.
Step 2
Prototype through the API
Use the documented Mercury 2.5 model identifier in an isolated test environment.
Step 3
Measure end to end
Track quality, tool-call correctness, P50/P99 latency, output size, and spend for real prompts.
Step 4
Plan for price change
Model cost at both current promotional and published list rates before committing a budget.
Cost
Mercury 2.5 pricing and free plan
Inception currently advertises an 80% promotional rate: $0.04 input and $0.15 output per million tokens for Mercury 2.5. Its regular displayed rates are $0.20 input and $0.75 output; cached input has separate lower rates. Enterprise hosting is quoted.
Promotional API
$0.04 input / $0.15 output per 1M tokens
Current 80%-off published offer.
- Cached input $0.004 per million tokens
- Promotion may end
- Available via Inception API and partners
Published list API
$0.20 input / $0.75 output per 1M tokens
Rate shown before the promotional discount.
- Cached input $0.02 per million tokens
- 260K context listed
- Use for conservative budgeting
Enterprise
Contact sales
Dedicated capacity and controls for larger deployments.
- Autoscaling options
- Compliance controls
- Configurable retention
Pricing checked . Check current pricing at the source ↗
Assessment
Mercury 2.5 strengths and limitations
Where it stands out
- Strong published latency and throughput positioning for repeated model calls.
- Tool calls and structured output support real application workflows.
- Current promotional unit price is low, with list rate visible for comparison.
What to consider
- Throughput and customer case studies are vendor-reported.
- Promotion timing and partner pricing can differ.
- A fast model may still need a stronger model for difficult or high-stakes reasoning steps.
Compare
Mercury 2.5 alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Consumer
GPT-6 Astra
Consider a frontier model when difficult reasoning matters more than per-call speed.
Explore GPT-6 Astra →Coding
Together AI
Consider a multi-model inference platform rather than one model family.
Explore Together AI →Coding
Google AI Studio
Consider Google’s developer environment for model experimentation.
Explore Google AI Studio →Questions
Mercury 2.5 FAQs
What is a diffusion language model?
It is an alternative generation architecture; Inception uses it to prioritize fast text output in Mercury 2.5.
What is the current rate?
The model page shows a promotional $0.04 input and $0.15 output per million tokens, 80% below displayed list rates.
Does it support agent tools?
The release lists parallel tool calls and schema-aligned JSON output.
Is 1,107 tokens per second guaranteed?
No. That is Inception’s reported benchmark on specified hardware, not a guaranteed application-level speed.
Bottom line
Our Mercury 2.5 verdict
Mercury 2.5 is a strong candidate for high-volume helper calls where latency compounds. Its discounted price is appealing, but the sound decision comes from an end-to-end workload test and a budget that still works at the displayed list rate.
Visit Mercury 2.5 website ↗Similar Miscellaneous Tools
View all Miscellaneous AI tools →
Muse Voice Transcribe
Muse Voice Transcribe - Meta's live speech-to-text that tracks 20+ speakers and mid-sentence language switches
GPT-Live-1
OpenAI's voice model that listens while talking, now for developers
Gemini Omni 1.1 Flash
Google's video model update with scene extensions and 4K upscaling
Atria Dawn Preview
Shanghai AI Lab's open agentic model for long research tasks
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.