Reported News
LongCat-2.5: a 1.6T-parameter multimodal agent model, live on the LongCat API at $0.30/$1.20
LongCat-2.5 is a native multimodal model for long-horizon agent work; its announcement says it has 1.6 trillion total parameters with about 48 billion active per token. The LongCat API change log lists it as available on 2026-11-02, with a 1M-token context and limited-time pricing of $0.30 input and $1.20 output per 1M tokens.
- Changes reported
- 6
- listed below, each from a source we read
- We tested
- 0
- we did not run it ourselves
- Gaps we flag
- 6
- things we could not verify
- Sources
- 4
- 4 primary, linked below
What changed
| # | What the source says | Status |
|---|---|---|
| 01 | The announcement describes LongCat-2.5 as a native multimodal model with 1.6 trillion total parameters and about 48 billion activated per token, building on LongCat-2.0. | Reported |
| 02 | Images are turned into discrete visual tokens so one backbone handles image and text input; the team pitches this at visual coding, creative design and 3D tasks. | Reported |
| 03 | The API docs give LongCat-2.5 a 1M-token context window and a 128K-token maximum output, through both OpenAI-format and Anthropic-format endpoints. | Reported |
| 04 | Pay-as-you-go pricing is listed as a limited-time discount: $0.30 uncached input, $0.006 cached input and $1.20 output per 1M tokens. | Reported |
| 05 | In the vendor's own table, LongCat-2.5 scores 88.4 on Terminal Bench v2.1 (LongCat-2.0: 70.8) and 91.9 on BrowseComp; competitor figures there are partly self-run. | Reported |
| 06 | The team reports 4-bit (W4A4) inference raised prefill throughput 43.3% and decode throughput 42.77% versus its INT8 baseline in online stress tests. | Reported |
reported by the sourcemeasured or run by usnot verified
Primary source: LongCat blog: Introducing LongCat-2.5. We are the fast, independent index; the vendor’s own page is the source of record for this release.
In the source’s words
“LongCat-2.5: 1M token context window with a maximum output length of 128K tokens”
“Compatible with Claude Code, Codex, Hermes, and other agent harnesses, with flexible integration into your development tools and workflows.”
“Pay-As-You-Go currently supports LongCat-2.5.”
What we have not verified
- We did not run this release ourselves.
- All benchmark scores are the vendor's own; the announcement says scores without an asterisk for other models were self-run and are indicative only.
- The announcement page is rendered by JavaScript, so its figures were read in a rendered copy and are not checked by our automated quote checker; the quotes above come from the static API docs.
- We found no open weights for LongCat-2.5 on the team's Hugging Face organisation at publication time, and the announcement we read does not say whether weights will be released.
- The pricing page calls the prices a limited-time discount and does not give an end date or the regular price.
- The change log and announcement give dates only, not a time of day, so the source publish time is left blank.
This brief is Reported: written from the sources below, not from running the release ourselves.
Sources
- LongCat blog: Introducing LongCat-2.5primary source · fetched 2026-11-02
- LongCat API Platform change logprimary source · fetched 2026-11-02
- LongCat API Platform quick start docsprimary source · fetched 2026-11-02
- LongCat API Platform pricing: LongCat-2.5primary source · fetched 2026-11-02
Related
Spotted an error? See our corrections policy or email the address on the contact page. Markdown version for agents: /news/longcat-2-5-native-multimodal-agentic-model.md
Want these briefs in your Google Top Stories? Add Agentic Weekly as a preferred source on Google.
The Weekly Verdict · free
5 new-model use cases. Every Monday.
A free email every Monday: the week's few big AI releases, then 5 short use cases for the new models, each with a prompt you can paste.
Free. Double opt-in. One click to unsubscribe.