Kimi K3.1 Pro API: Up to 70% Off, Pricing & Release Timeline

Kimi K3.1 release date: from teaser to Pro launch

The Pro launch follows the October 23 official demonstration; by the October 24 check, the launch blog, model list and API guide confirm public availability. Earlier teaser dates describe earlier signals, not the date every account received access.

The earlier internal-testing screenshot remains as historical context:

The embedded Lentils post in that screenshot is dated August 13. It is older discussion, not a separate Pro launch announcement.

Kimi K3.1 Pro specs: context, modalities and reasoning

Kimi describes Pro as the first model in the K3.1 series and a 3T-class model post-trained from Kimi K3, with substantially larger post-training compute. The documentation now supplies model-specific limits; these are no longer inferred from a short demo.

A 1M context is a shared request budget, not a promise of simultaneous 1M input plus 1M output. The exact output allowance also depends on the serving provider. The official model list calls Pro 3T-class; this article does not promote a separately unverified active-parameter count into a confirmed specification.

Kimi K3.1 Pro benchmarks: coding, 3D and finance

The following results come from Kimi's launch blog and its benchmark figures, not an APIMaster evaluation. Kimi says its Pro evaluations generally use reasoning_effort=max, temperature 1.0 and top-p 0.95 unless otherwise specified. Harnesses and evaluation setups differ by benchmark.

The capabilities behind those results are concrete: full-project builds in Unity, Unreal Engine and Godot; parametric Blender and CAD modeling through scripts; and financial modeling with linked formulas and source-grounded research. Kimi's internal finance scores improve by 7.2 points for modeling and 11.6 points for research over K3.

The drop from 23.4% to 8.3% is approximately 65% relative reduction on that hallucination stress test. It is not a measured reduction in every user's daily error rate. Likewise, one strong coding score does not establish that Pro wins every task or consistently beats Opus.

Kimi K3.1 Pro vs Highspeed: what changes?

Kimi describes Highspeed as offering the same intelligence at 2–3× the speed. It charges twice the uncached-input and output rates, while cached-input pricing is the same.

For batch analysis and long coding jobs, Pro's lower token rates may matter more. For interactive agents, Highspeed may reduce waiting time, but compare time to first token, completed-task latency and total tokens on your own workload. A relative speed claim is not a universal tokens-per-second guarantee.

K3.1 Pro is also distinct from the older K2.8 Preview and its kimi-for-coding access surface. In Kimi Code, use /model and select the released Pro model rather than assuming an old alias maps to it.

Kimi K3.1 Pro API pricing: input, output and caching

The official pricing page lists the following USD rates per 1M tokens, excluding applicable taxes. Kimi's billing FAQ says Pro uses flat token rates, with no context-length price tiers.

Pro's standard input and output rates are each 33.3% below K3's official rates. For illustration, 100,000 input tokens plus 10,000 output tokens at the base non-cached rates cost $0.30 on Pro, $0.60 on Highspeed and $0.45 on K3. This arithmetic excludes any separate cache-write items, external tools, taxes and repeated agent calls.

Kimi automatically enables context caching for repeated prefixes, with a default 5-minute TTL. Cache writes are separate billing items; cached input uses the cache-hit rate. Check usage details and TTL settings rather than estimating a long agent run from prompt size alone.

Official API pricing is separate from a Kimi app or Kimi Code subscription. The API guide requires a successful top-up of at least $1 to unlock the flagship model; rate limits depend on the account tier.

APIMaster Kimi prices: compare the current route before buying

The new Pro market card currently lists 2 active public routes from $0.60 / $3 to $2 / $10 per 1M input / output tokens. The Highspeed card has 0. These October 24 marketplace results supersede the earlier “no K3.1 routes” statement.

The lowest Pro route offers up to 70% off the official input and output rates. This compares base token rates; cache writes, tool charges and provider-specific limits require their own checks. We have not run a paid Pro request for this article.

For existing-model workloads, Kimi K3 has 7 active routes, ranging from $1.938 / $9.69 to $3 / $15, or up to approximately 35% off official K3 pricing. K2.7 Code has 5 routes from $0.7125 / $3 to $0.95 / $4, up to 25% off. K2.6 has 2 routes at $0.95 / $4, with no discount.

These are different models with their own route prices. Compare Pro with its own official $2 / $10 baseline, and K3 with $3 / $15. Kimi also states that K2-series models will close to new users on October 31, with retirement planned by the end of December 2026; that is Kimi's official platform schedule, not a verified APIMaster shutdown date.

How to use the Kimi K3.1 Pro API

The official quickstart uses an OpenAI-compatible client and the Moonshot endpoint. Install the OpenAI SDK, set your own MOONSHOT_API_KEY, and use the exact released model ID:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["MOONSHOT_API_KEY"],
    base_url="https://api.moonshot.ai/v1",
)

response = client.chat.completions.create(
    model="kimi-k3.1-pro",
    reasoning_effort="high",
    max_completion_tokens=4096,
    messages=[
        {"role": "user", "content": "Write a Python function to validate an IPv4 address."}
    ],
)
print(response.choices[0].message.content)

To use the official high-speed variant, change the model to kimi-k3.1-pro-highspeed. Pro always has thinking enabled; choose low, high or max to tune effort. In multi-turn or tool-call workflows, preserve the complete assistant message, including reasoning fields returned by the service, in the following request as documented.

For APIMaster, use an APIMaster key, base URL https://apimaster.ai/v1, and a model listed on the chosen live card. Confirm the Pro price before changing a production client; kimi-k3 remains a separate available model. Earlier sightings such as kimi-k3-1 and kimi-k3.1 are historical identifiers, not substitutes for the documented kimi-k3.1-pro.

FAQ

Is Kimi K3.1 Pro released, and when?

Yes. The October 23 official video was followed by the Pro launch rollout, and the official blog and API documentation checked on October 24, 2026 confirm Pro and Highspeed availability across Kimi, Work, Code and the API. This supersedes the earlier preview-only status.

What is the Kimi K3.1 Pro API price?

Official Pro rates are $2 input / $10 output per 1M tokens, with $0.20 cached input. Highspeed is $4 / $20, also with $0.20 cached input. Cache writes have separate TTL-based rates. The official rate is not the same as an APIMaster route quote.

What are the exact API model IDs?

Use kimi-k3.1-pro or kimi-k3.1-pro-highspeed on the documented official endpoint. The earlier reported kimi-k3-1 and tracking label kimi-k3.1 are not the released IDs listed by Kimi.

Is Kimi K3.1 Pro open weight already?

Kimi describes Pro as an open-weight model and commits to releasing the full weights by November 4, 2026. At this October 24 check, that is a planned checkpoint release; API availability alone does not establish a downloadable checkpoint or its final license.

How is Highspeed different from Pro?

Kimi says it has the same intelligence and runs 2–3× faster. Its uncached-input and output prices are twice Pro's, while cached input remains $0.20 per 1M tokens. APIMaster has no public Highspeed route in this check.

Can I use Kimi K3.1 Pro through APIMaster now?

The marketplace returns 2 active public Pro routes on October 24, from $0.60 input / $3 output to $2 / $10 per 1M tokens. The lowest is approximately 70% below the official base rates. No paid request was tested for this article; check the live card and final quote. Highspeed has no active public route in this check.

Sources and verification date

• Kimi K3.1 Pro launch blog: official availability, capabilities, benchmark values and conditions, Highspeed speed claim, and November 4 weight-release commitment; checked October 24, 2026.
• Kimi K3.1 Pro API quickstart: model IDs, access requirements, reasoning controls, multimodal input and completion settings.
• Official model list, token pricing, account and billing, and context caching: model availability, prices, cache-write billing and migration notes.
• October 23 official video: earlier public demonstration. ChainCatcher's October 23 report supplies the attributed late-September and October 21 sightings; original API/configuration records were not independently verified.
• lyra (@lyraxana) and Kimi's verified Zhihu account: historical internal-testing screenshot and September 19 teaser retained from this tracker.
• Kimi Code model documentation: product-specific model selection.
• APIMaster public marketplace API, checked 2026-10-24, for kimi-k3.1-pro, kimi-k3.1-pro-highspeed, kimi-k3, kimi-k2.7-code and kimi-k2.6; only active public routes counted. Route availability and prices are a timestamped snapshot.

Choose a Kimi API route with a clear price

The newly released Kimi K3.1 Pro starts at $0.60 / $3, approximately 70% below its official $2 / $10 base rates in this check. The older K3 card starts at $1.938 / $9.69; choose the model and route for your workload.

Register with APIMaster, create an API key in the console, and configure https://apimaster.ai/v1 with your chosen documented model ID. Use kimi-k3 for the K3 route; select kimi-k3.1-pro only after reviewing its current price and supported limits.