Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Inception / Mercury

Mercury 2.5

Proprietary diffusion language model from Inception.

Language models (LLMs)
availableClosed weightsText generationCoding
Official documentation (opens in a new tab)Reviewed 15 Oct 2026
13 of 18 specification fields collected
Context window260K260,000 tokens
Maximum output65.5K65,536 tokens
Input / 1M tokens$0.04Inception API
Output / 1M tokens$0.15USD · observed 15 Oct 2026

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
260,000 tokensSource for Context / input limit (opens in a new tab)

Combined context budget, not an additional separate input allowance.

Maximum input tokens
Not collected

No exact model-specific contract established in the reviewed evidence.

Maximum output
65,536 tokensSource for Maximum output (opens in a new tab)
Parameters
Not collected

No exact model-specific contract established in the reviewed evidence.

Release date
2026-09-08Source for Release date (opens in a new tab)
Knowledge cutoff
Not collected

No exact model-specific contract established in the reviewed evidence.

Reasoning
SupportedSource for Reasoning (opens in a new tab)
Tool calling
SupportedSource for Tool calling (opens in a new tab)
Structured output
SupportedSource for Structured output (opens in a new tab)
Streaming
SupportedSource for Streaming (opens in a new tab)
Batch processing
Not collected

No exact model-specific contract established in the reviewed evidence.

Fine-tuning
Not collected

No exact model-specific contract established in the reviewed evidence.

Context caching
SupportedSource for Context caching (opens in a new tab)

Automatic prefix caching; no documented TTL guarantee.

Input modalities
TextSource for Input modalities (opens in a new tab)
Output modalities
TextSource for Output modalities (opens in a new tab)
Architecture
Diffusion language model using parallel iterative refinement.Source for Architecture (opens in a new tab)

Exact parameter count and layer architecture are not published here.

Generation controls
Reasoning: instant, low, medium or high; default medium. Temperature 0.5–1; out-of-range values reset with a warning. No seed parameter.Source for Generation controls (opens in a new tab)

Mercury 2 default temperature 0.75; Mercury 2.5 default 1.

Request quotas
Account tiers: Free 1,000 RPM / 1M input TPM / 100K output TPM; PAYG 3,000 RPM / 3M input TPM / 300K output TPM; enterprise custom.Source for Request quotas (opens in a new tab)

Platform quotas, not a per-model latency or throughput guarantee.

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

Inception API

Provider source (opens in a new tab)
API model ID
mercury-2.5
Access
API-key access. Public discovery metadata was read without authentication; no billed inference was performed.
Regions
US is listed by live discovery; this is not a complete contractual residency guarantee.
API endpoint
POST /v1/chat/completionsSource for API endpoint (opens in a new tab)
Fine-tuning
Not supportedSource for Fine-tuning (opens in a new tab)

No public self-serve fine-tuning endpoint; custom enterprise requests are separate.

Generation controls
Default output budget 16,384; with high reasoning, default 65,536. Maximum 65,536.Source for Generation controls (opens in a new tab)

max_completion_tokens includes reasoning and takes precedence over deprecated max_tokens.

INPUT / 1M TOKENS$0.04
OUTPUT / 1M TOKENS$0.15

Advertised promotional snapshot on October 15, 2026. End date not published; verify current account pricing. Billed output includes reasoning. Cached input price: $0.004 / 1M tokens (base tier).

USD · observed 15 Oct 2026 · Pricing evidence (opens in a new tab)
Published billing units · original currency · observed 15 Oct 2026
CategoryUsageRateUnit
Cache chargesAdditional cache write rate1$0.001M tokens
Undiscounted reference pricesInput2$0.201M tokens
Undiscounted reference pricesOutput2$0.751M tokens
Undiscounted reference pricesCached input2$0.021M tokens

Rate conditions

  1. The discovery response lists input_cache_writes as zero. This does not remove ordinary input charges or guarantee cache reuse.
  2. Crossed-out regular rate in the provider pricing table, not the advertised promotional rate. No promotion end date is published.
Pricing source (opens in a new tab)Pricing source (opens in a new tab)

Data & enterprise terms

Terms permit using submissions for model training unless the user disables “Improve the model for everyone” in API Platform settings. The privacy policy uses purpose-dependent retention, not a fixed universal period. The FAQ says custom zero-retention arrangements are handled case by case; enterprise agreements may differ.

Read the service policy (opens in a new tab)

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

Inception API

mercury-2.5
Documented
Support evidence (opens in a new tab)
How it works
Automatic prefix caching.Source for How it works (opens in a new tab)
Minimum reusable input
Not collected

Minimum prefix length not documented.

Lifetime & refresh
No documented TTL; pinning, invalidation and cache inspection are not exposed.Source for Lifetime & refresh (opens in a new tab)
How to enable it
Enabled automatically; no request parameter required.Source for How to enable it (opens in a new tab)
Usage fields
prompt_tokens includes cached_tokens; completion_tokens includes reasoning_tokens. Do not add nested counts again. Streaming usage requires stream_options.include_usage.Source for Usage fields (opens in a new tab)
Billing rules
Cache reads use the model-specific cached-input rate. Reasoning tokens are billed as output.Source for Billing rules (opens in a new tab)
Limits & data boundaries
Cache lifetime is not a customer-data-retention promise.Source for Limits & data boundaries (opens in a new tab)
Cache reads, writes & storage
ChargeApplies toRate / unit
Cache read1Text tokens · 80% promotion · end date unpublished$0.004 / 1M tokensPricing evidence (opens in a new tab)
Additional cache write rate2Cache charges$0.00 / 1M tokensPricing evidence (opens in a new tab)
Cached input3Undiscounted reference prices$0.02 / 1M tokensPricing evidence (opens in a new tab)

Rate conditions

  1. Advertised promotional snapshot on October 15, 2026. End date not published; verify current account pricing. Billed output includes reasoning.
  2. The discovery response lists input_cache_writes as zero. This does not remove ordinary input charges or guarantee cache reuse.
  3. Crossed-out regular rate in the provider pricing table, not the advertised promotional rate. No promotion end date is published.

Observed 15 Oct 2026. Write rates and storage fees are different charges; use the billing rules above.

Caching reviewed 15 Oct 2026

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Inception’s own Trust Center describes the Mercury family as proprietary. Public API and enterprise deployment access do not establish public weights.

Classification evidence (opens in a new tab)

Lifecycle dates

Generation announced8 Sept 2026Source (opens in a new tab)

Sources reviewed