What is Mercury 2?
Mercury 2 is a reasoning model from Inception that writes by diffusion: it sketches a whole reply, then refines many tokens at once over a few passes instead of adding one token at a time. Inception measured 1,009 tokens a second on NVIDIA Blackwell GPUs and pitched it for voice agents, search and coding tools, where every model call adds delay. It reads and writes text only.
Inception has closed it to new users. Customers with earlier usage keep access until November 15, 2026, enterprise customers stay covered by their agreements, and everyone else is pointed to Mercury 3, which reads 1M tokens and costs $0.20 / $0.70 per million tokens.
Is Mercury 2 still available?
Only to existing customers until November 15, 2026, and to enterprise customers under their agreements; it is closed to new users.
- New users
- Cannot use Mercury 2; new Inception API keys start on Mercury 3.
- Existing customers
- Accounts with usage before Mercury 3 launched keep access until November 15, 2026.
- Enterprise customers
- Stay supported under their agreements, with migration help from their Inception representative.
- Moving over
- Mercury 3 runs on the same chat completions endpoint, so switching means changing the model ID to mercury-3 and retesting.
Who is Mercury 2 best for?
Mercury 2 is best for existing Inception customers who need its fast reasoning calls to keep running until they finish moving to Mercury 3.
Existing customers mid-migration
Accounts that used Mercury 2 before Mercury 3 launched keep access until November 15, 2026, long enough to compare outputs side by side before switching.
Enterprise contracts
Inception keeps supporting enterprise customers under their agreements, so a contracted deployment does not stop at the public cutoff.
A baseline for Mercury 3 tests
Mercury 3 uses the same chat completions endpoint and request format, so the same prompts can run on both models while you check quality and speed.
What's included with Mercury 2?
Four reasoning effort levels, tool calling and structured output, an OpenAI-compatible API and automatic prompt caching.
- Reasoning effort
- Set instant, low, medium or high per request to trade thinking time for speed; medium is the default.
- Tool calling and structured output
- It calls functions natively and returns JSON that follows a schema you supply.
- OpenAI-compatible API
- Point any OpenAI-style client or agent framework at Inception's endpoint and change the model name.
- Automatic prompt caching
- Repeated prompt prefixes are cached with no setting and billed at $0.025 per million tokens.
How much does Mercury 2 cost?
Mercury 2 costs $0.25 per million input tokens and $0.75 per million output tokens on Inception's API, but it is closed to new users and existing access ends November 15, 2026; Mercury 3 replaces it at $0.20 / $0.70.
- Cached input at $0.025 per million tokens
- Closed to new users
- Access ends November 15, 2026
- Supported under enterprise agreements
- Custom rate limits, SLAs and volume pricing
- Migration help to Mercury 3 from Inception
Inception has closed Mercury 2 to new users: customers who used it before Mercury 3 launched keep access until November 15, 2026, and enterprise customers stay supported under their agreements. Its replacement, Mercury 3, lists $0.20 per million input tokens, $0.02 cached and $0.70 output, with a 1M-token context, and new API keys come with 100 million free tokens to try it. Full pricing on inceptionlabs.ai. We checked these prices on .
How do readers rate Mercury 2?
No one has rated Mercury 2 yet. Be the first.
Sign in to rate. One rating per account.
Mercury 2 pros and cons
It was very fast and cheap for a reasoning model, but users report more factual slips than larger models, it is closed to new users, and existing access ends November 15, 2026.
Pros
- Very fast in practice: users on Hacker News and Reddit were struck by its speed, in line with Inception's figure of over 1,000 tokens a second on standard NVIDIA GPUs.
- Cheap for a reasoning model at $0.25 / $0.75 per million tokens, which Inception put at about half a cent per minute for a typical voice agent.
- Beat GPT-4.1 in Inception's tests at medium effort, by 27 points on IFBench and 24 points on Tau3Bench Telecom, while still answering faster.
- Users who ran it as a voice agent's brain or as an LLM judge found it good enough for those quick jobs.
Cons
- Closed to new users, and existing access ends November 15, 2026 for everyone outside an enterprise agreement; Mercury 3 is the replacement.
- Makes more factual slips than larger models: Hacker News and Reddit users caught it misspelling names, inventing facts and failing simple trick questions, and some would rather wait for a slower, more accurate model in coding agents.
- Superseded twice: Mercury 3 reads 1M tokens instead of 128K, costs $0.20 / $0.70 per million tokens instead of $0.25 / $0.75, and Inception puts it on par with GPT-6 Luna and Claude Haiku 5.5.
- Closed and text only: no downloadable weights, no image input, no fine-tuning through the public API, and temperature is locked between 0.5 and 1.0 with no seed setting.
What are the best Mercury 2 alternatives?
Mercury 3 as the direct replacement, Mercury Voice for voice agents, and Gemini 3.1 Flash-Lite for multimodal input and a free tier.
Mercury 3FreemiumChoose Mercury 3 for anything you run on Mercury 2 today: it is Inception's replacement on the same endpoint, with a 1M-token context, up to 131,072 output tokens and $0.20 / $0.70 per million tokens.
Mercury VoicePaidChoose it when Mercury 2 was the brain of a voice agent: Inception tuned it for live calls, with the first token in under 170 ms, at $0.40 / $1.50 per million tokens.
Gemini 3.1 Flash-LiteFreeChoose it when you need image, video, audio or PDF input or a 1M-token context; it costs $0.25 / $1.50 per million tokens and has a free tier.
Questions about Mercury 2
What is Mercury 2?
Mercury 2 is Inception's diffusion reasoning model, which drafts whole passages in parallel to reply at over 1,000 tokens a second; it is closed to new users, and existing access ends November 15, 2026.
Who is Mercury 2 for?
Mercury 2 is best for existing Inception customers who need its fast reasoning calls to keep running until they finish moving to Mercury 3.
Is Mercury 2 free?
No. New Inception API keys come with 100 million free tokens, but new accounts cannot use Mercury 2 because it is closed to new users. Existing customers pay $0.25 per million input tokens and $0.75 per million output tokens until access ends November 15, 2026.
How much does Mercury 2 cost?
Mercury 2's paid plans are Developer (existing customers) at $0.25 / $0.75 per million input / output tokens.
What are the best Mercury 2 alternatives?
The closest alternatives on Recaply are Mercury 3, Mercury Voice and Gemini 3.1 Flash-Lite.
Where can I try Mercury 2?
On its official website, inceptionlabs.ai.
Is Mercury 2 being shut down?
Yes. Inception has closed Mercury 2 to new users, and existing customers keep access only until November 15, 2026; enterprise customers stay supported under their agreements. Inception points everyone else to Mercury 3 and offers migration help through its docs and account team.
How does Mercury 2 compare with Mercury 3?
Mercury 3 is Inception's replacement: it reads 1M tokens instead of 128K, writes up to 131,072 tokens per reply instead of 50,000, and costs $0.20 / $0.70 per million tokens against $0.25 / $0.75. Inception rates its quality on par with GPT-6 Luna and Claude Haiku 5.5.
How is Mercury 2 so fast?
Mercury 2 is a diffusion model: instead of writing one token per step, it refines many tokens of a draft in each pass, so the GPU does more useful work per step. Inception measured 1,009 tokens a second on NVIDIA Blackwell GPUs.
Can Mercury 2 read images?
No. Inception's documentation says Mercury 2 accepts text input only, with no image input or image generation. For pictures, PDFs or audio you need a multimodal model such as Gemini 3.1 Flash-Lite.
Our verdict
Mercury 2 was a fast, cheap reasoning model, but it is closed to new users and existing access ends November 15, 2026.
Inception has replaced it with Mercury 3, which reads 1M tokens instead of 128K and costs less per token, and only enterprise agreements keep Mercury 2 running past the cutoff.
- Skip it for
- Any new project, since new accounts cannot use it, and work where a wrong fact costs more than a slow answer.
- Check first
- If you run it now, move your prompts to Mercury 3 and retest them before November 15, 2026; on Inception's API the switch is a change of model ID.
DiffusionGemma
SubQ
Muse Spark
Ring-2.6-1T
Qwen3-Max-Thinking