MiniMax has launched M3, which it bills as the first open-weight model to combine frontier coding and agentic capabilities, a million-token context window built on its proprietary MiniMax Sparse Attention architecture, and native multimodality.

Cystic echinococcosis (CE), caused by the metacestode of Echinococcus granulosus sensu lato, is a zoonosis that remains endemic across the Middle East, the Mediterranean basin, North Africa and Sou...

1.6万亿总参大模型，训练全程由国产芯片完成

1.6万亿总参大模型，训练全程由国产芯片完成

Head and neck squamous cell carcinoma (HNSCC) represents a significant public health challenge. Its treatment is characterized by high anatomical complexity and a critical need for functional preservation of speech, swallowing, and respiration. While multimodality care, including surgery, radiotherapy, and systemic therapy, has substantially improved patient survival, there are still high rates of life-altering toxicities and low response rates to immunotherapy. Current clinical decision-making largely relies on “digital snapshots”—static representations derived from population-based statistics and time point–specific imaging—which usually fail to account for the rapid anatomical and biological change during the course of treatment. Digital twins hold great promise in accelerating scientific breakthroughs and transforming oncology treatment and precision medicine through the creation of high-fidelity virtual patient representations that integrate real-time biological, anatomical, and clinical data. In this viewpoint, we propose 3 interconnected digital twins to establish a conceptual framework for the HNSCC digital twin. These digital twins include the anatomical twin, using virtual surgical planning and augmented reality for geometric precision; the dosimetric twin, using daily imaging and synthetic computed tomography generation for automated adaptive radiotherapy; and the biological twin, integrating deep phenotyping, radiomics, and mechanistic omics to simulate individualized therapeutic trajectories. Beyond acute care, we explore the clinical utility of digital twins in navigating the “de-escalation dilemma” for human papillomavirus–associated disease and improving survivorship through digital phenotyping. With the advent of longitudinal data from medical-grade Internet-of-Things sensors and circulating tumor DNA kinetics, digital twins can serve as active sentinels, detecting functional decline or subclinical events months before clinical manifestation. Furthermore, the implementation of synthetic control arms via virtual populations offers an avenue to accelerate clinical trials and address ethical issues in rare molecular subtypes. Despite this potential, substantial barriers to implementation remain, including profound anatomical instability during treatment, the “black box” nature of deep learning algorithms, and the challenges of multiscale data integration. We provide a technical roadmap for the development of “morphing” digital twins—dynamic systems capable of continuous geometric and biological auditing. By bridging the gap between the macroscopic world of radiation physics and the microscopic world of cellular evolution, digital twins may help shift the management of HNSCC from population-based statistical probability toward more individualized, biologically informed decision-making, with the goal of optimizing HNSCC management while preserving essential patient functions.

> **Attributed compile + one small offline run on my box** > > **Primary source:** Introducing LongCat-2.5 — Native Multimodality for Agentic Tasks by **Meituan / LongCat team** , published **2026-11-1**. > > **Pricing / billing:** LongCat-2.5 pricing and API Pay-As-You-Go (limited-time discounted rates; platform billing authoritative). > > **API:** Chat Completions + Platform Quick Start — OpenAI-compat `https://api.longcat.chat/openai`, Anthropic-compat `https://api.longcat.chat/anthropic`, model id **`LongCat-2.5`**. > > **Secondary CN press:** 美团发布LongCat-2.5大模型，“国模国芯”实现新突破 (Meituan / 经济观察报 framing, **2026-11-02**) for the “scores between Claude-Opus-4.8 and Claude-Opus-5” claim and product/API availability. > > Benchmarks and prices below are **vendor-reported** unless I say otherwise. This box did **not** call chat completions. Offline preflight = public HTTP scrapes + list-price arithmetic only. Meituan shipped **LongCat-2.5** on **2026-11-1** (blog) with wide coverage on **2026-11-2** : a **1.6T** -parameter MoE (~**48B** activated per token) pitched as a **native multimodal** model for **long-horizon agentic** work. For builders, the useful story is not “another 万亿 MoE.” It is four concrete checks: (1) **OpenAI- and Anthropic-compatible** endpoints with a single model string, (2) a **`thinking`** object that **forces temperature** (1.0 on / 0.7 off), (3) limited-time list price **$0.30 / $0.006 cached / $1.20** per 1M tokens, and (4) **vendor** benches that Meituan’s own CN framing places **between Claude Opus 4.8 and Opus 5** — which is a reason to **run your golden eval** , not a reason to merge their table into your prod decision. Try-it surface is **longcat.ai** ; developers get a prepaid Pay-As-You-Go API the same week. That is enough to A/B against a current default **without** waiting for open weights — and also enough to over-trust a vendor dashboard if you skip your own tickets. ## What it is (short) LongCat-2.5 builds on **LongCat-2.0** with stronger native visual understanding and agentic coding (vendor). Visual signals are represented as **multi-level discrete tokens** (RVQ over a Semantic-and-Aligned Encoder / SAE-style path), so vision and text share a unified token stream. Training emphasis is **long-horizon RL** : more long trajectories in long-sequence training, failure trajectories turned into verifiable tasks, mixed training across harnesses. Infra notes (for context only): long-context training and **W4A4** inference on domestic AI accelerators; vendor online stress vs W8A8 INT8 cites prefill **+43.3%** , decode **+42.77%** , TTFT **−15.91%** , TPOT **−19.72%** , GMM kernel ~**1.73×**. That is serving math, not permission to skip quality gates. Site / docs language also says the model works across **Claude Code** , **Codex** , and **Hermes** harnesses. Treat that as **compatibility marketing** until you wire one harness and measure your own tasks. Blog eval footnotes themselves show different harnesses per bench (Claude Code on several coding rows, Codex on SkillsBench, custom browse harness on BrowseComp) — another reminder that “works across” is not “one config wins every leaderboard.” Skip the deep train–inference / supernode digression for day-one integration. Remember only: long context (**1M**) is a first-class claim, multimodal is native (not a bolted vision sidecar in the marketing story), and agentic RL is the capability bet. ## How to call it: dual endpoints + the temperature lock Platform docs advertise two bases: Format | Base URL | Typical path ---|---|--- OpenAI-compatible | `https://api.longcat.chat/openai` | `/v1/chat/completions` Anthropic-compatible | `https://api.longcat.chat/anthropic` | `/v1/messages` **Model id:** `LongCat-2.5`. Context window: **1M** tokens; max output: **128K**. Default `top_p` on chat docs: **0.95**. Thinking is an object, not a free-form string: "thinking": { "type": "enabled" } or `{ "type": "disabled" }`. **Trap:** chat docs say **temperature is fixed at`1.0` with thinking enabled and `0.7` with thinking disabled**; other values are **overridden**. If your client still sends `temperature: 0` from an older Anthropic/OpenAI template, you are not getting determinism — you are getting the forced default for the thinking mode you picked. Eval note on the blog: LongCat-2.5 rows use `enable_thinking=on`, `temperature=1.0`, `top_p=0.95` unless specified; other models marked `*` are public reports. OpenAI SDK shape is the usual `base_url="https://api.longcat.chat/openai"` + Bearer key. Anthropic SDK path uses `https://api.longcat.chat/anthropic` with the docs’ header pattern (`Authorization: Bearer …` plus Anthropic-style body). Pick **one** client stack for the A/B; do not flip formats mid-eval or you will confuse tool-schema bugs with model bugs. Builder rule: set **`thinking` explicitly per route**, assume temperature will be **clamped** , and pin the base URL + model string in config — do not guess. If you need lower temperature for classification, you likely want **thinking disabled** (and you still get **0.7** , not 0.0). Document that in the route table so the next engineer does not “fix” sampling with a dead knob. ## Vendor benches (label as vendor) From the LongCat-2.5 blog metrics table (selected high-signal rows). **LongCat-2.5 only** below; peer columns exist on the source page — do not invent third-party re-scores here. Benchmark (vendor table) | LongCat-2.5 | vs LongCat-2.0 (same table) ---|---|--- Terminal Bench v2.1 | **88.4** | 70.8 Terminal Bench v4.0 | **38.5** | 3.0 DeepSWE v1.1 | **64.6** | 9.7 PaperBench | **82.1** | 45.2 BrowseComp | **91.9** | 79.9 AutomationBench v1.0.6 | **51.8** | 15.8 HLE w/tools | **63.1** | 37.5 Vision2Web | **78.6** | — 3DCodeBench Text→3D | **55.4** | — Meituan’s **2026-11-02** CN press frames results as sitting **between Claude-Opus-4.8 and Claude-Opus-5** on several mainstream benches (and exceeding Opus-5 on some). That is a **vendor / press claim** next to their own charts — useful as a map of where _they_ want you to look, useless as a substitute for your golden set. Same caution as this series always uses: do not paste “between Opus 4.8 and 5” into a slide that implies an independent Artificial Analysis or third-party harness unless that harness is cited. Jump vs 2.0 on Terminal Bench v4.0 (**3.0 → 38.5**) and DeepSWE (**9.7 → 64.6**) is the internal “agentic coding moved” signal. Still: those are **their** runs with **their** thinking/temp settings. Replicate on your repos before you retire the incumbent model string. ## List price: $0.30 / $0.006 cached / $1.20 (limited-time) Pay-As-You-Go currently lists **LongCat-2.5** at discounted (limited-time) rates: Item | USD / 1M tokens | CNY / 1M tokens ---|---|--- Uncached input | **$0.30** | **¥2** Cached input | **$0.006** | **¥0.04** Output | **$1.20** | **¥8** Prices are subject to change; platform UI + billing records win. Failed requests (**401 / 403 / 429 / 500**) are **not** billed; only **HTTP 200** is. ### List-price arithmetic (my assumptions — not a bill) Offline preflight used those three USD rates only. **Assumptions:** one trajectory = **80k input + 20k output** ; “80% cache hit” means 80% of _input_ tokens bill at the cached rate; no tax, retries, or tool-overhead tokens. Scenario | USD (list math) ---|--- 1 traj, no cache | **$0.0480** 1 traj, 80% input cache | **$0.0292** 40-turn loop, no cache | **$1.9200** 40-turn loop, 80% input cache | **$1.1674** 1M in + 1M out, no cache | **$1.5000** Cache is the swing factor on long agent loops. Thinking-on also tends to inflate **reasoning / output** tokens relative to thinking-off — price the knob and the rate card together, and remember temperature will not save you from that inflation. For a rough sanity check: at these list rates, **1M in + 1M out** is **$1.50** before cache; an 80k/20k turn with fat prefix reuse lands closer to **~$0.03** if the 80% cache assumption holds. Measure real `cached` vs `uncached` fields on usage before you put the middle rows in a forecast. CNY list (**¥2 / ¥0.04 / ¥8**) is on the same docs pages — use it if you bill in RMB; this post’s arithmetic sticks to USD for series consistency. ## Checklist before you swap 1. **Freeze a golden eval** on _your_ agent failures (repo patches, terminal tasks, browse/tool traces, visual coding if that is your product). Vendor TB / BrowseComp / Vision2Web are someone else’s harness. 2. **Pin base URL + model id** — `LongCat-2.5` on either `api.longcat.chat/openai` or `…/anthropic`. Confirm which SDK headers your client needs (Anthropic path wants the usual Anthropic-shaped request). 3. **Set`thinking` explicitly.** Assume **temperature is overridden** (1.0 on / 0.7 off). Do not rely on a silent `temperature: 0` from old templates. 4. **Price both columns:** list $0.30/$1.20/$0.006 _and_ expected output tokens with thinking on. Re-run the arithmetic with your real prompt sizes. Treat rates as **limited-time**. 5. **Harness claim ≠ drop-in.** If you use Claude Code / Codex / Hermes, prove one path with your tools and auth before you rewrite defaults. 6. **Multimodal only if you need it.** Discrete visual tokens are the product story; your golden set should include the visual tasks you actually ship. 7. **Billing hygiene:** prepaid balance; only HTTP 200 billed; do not forecast from failed 429 storms. 8. **Do not conflate** Meituan vendor rows, CN “between Opus 4.8 and 5” framing, and any third-party index you have not opened yourself. ## What I actually ran (offline) `/workspace/reputation/runs/31-longcat-2.5/` — 2026-11-02 America/Toronto: LONGCAT_API_KEY: NOT SET Chat completions: NOT ATTEMPTED HTTP 200: longcat.ai/blog/longcat-2.5/ HTTP 200: platform pricing + pay-as-you-go + chat + docs + homepage HTTP 200: meituan.com CN press (Opus 4.8–5 framing) Needle hits: $0.30 / $0.006 / $1.20, LongCat-2.5, openai/anthropic bases, thinking + temperature lock language, 1M/128K, Claude Code/Codex/Hermes List-price math (assumptions above): 1 traj no cache $0.0480 | 80% cache $0.0292 40-turn no cache $1.9200 | 80% cache $1.1674 1M+1M $1.5000 NOT tested: live chat; thinking/temp override live; harness wiring; Terminal Bench / DeepSWE / BrowseComp re-score; real cache hits; W4A4 latency Costs are arithmetic on LongCat’s stated limited-time list rates under assumptions I state. They are not a bill, not latency, and not quality. ## When LongCat-2.5 might become a default Only after a callable path and a golden eval win on _your_ tickets: 1. Long-horizon coding / agent loops where your current default fails and an OpenAI- or Anthropic-shaped client is already in the stack. 2. You have chosen **thinking on/off per route** , knowing temperature is **locked**. 3. Cost model survives realistic cache **and** thinking-on output inflation at $0.30/$1.20. 4. You verified the harness you actually run (not only the brochure list). 5. You can roll back without rewriting orchestration. If you are already deep in Anthropic- or OpenAI-shaped agent code, LongCat-2.5 is a **same-week** candidate for a side-by-side — not because the Opus framing is gospel, but because the integration surface is familiar and the list price is easy to model. If you need open weights, local GPUs, or a third-party index row before you trial anything, this drop is the wrong trigger; wait for a surface that matches that constraint. LongCat-2.5’s builder-relevant release is the **dual-compat API** , the **thinking/temperature lock** , the **$0.30/$1.20** limited-time card, and **vendor** benches framed near Opus 4.8–5. Treat those as labeled inputs, run your own eval, and only then change the production model string. YongBo Yu · Toronto AI engineer · https://yongbo-yu.vercel.app · GitHub YongBoYu1

Head and neck squamous cell carcinoma (HNSCC) represents a significant public health challenge. Its treatment is characterized by high anatomical complexity and a critical need for functional preservation of speech, swallowing, and respiration. While multimodality care, including surgery, radiotherapy, and systemic therapy, has substantially improved patient survival, there are still high rates of life-altering toxicities and low response rates to immunotherapy. Current clinical decision-making largely relies on “digital snapshots”—static representations derived from population-based statistics and time point–specific imaging—which usually fail to account for the rapid anatomical and biological change during the course of treatment. Digital twins hold great promise in accelerating scientific breakthroughs and transforming oncology treatment and precision medicine through the creation of high-fidelity virtual patient representations that integrate real-time biological, anatomical, and clinical data. In this viewpoint, we propose 3 interconnected digital twins to establish a conceptual framework for the HNSCC digital twin. These digital twins include the anatomical twin, using virtual surgical planning and augmented reality for geometric precision; the dosimetric twin, using daily imaging and synthetic computed tomography generation for automated adaptive radiotherapy; and the biological twin, integrating deep phenotyping, radiomics, and mechanistic omics to simulate individualized therapeutic trajectories. Beyond acute care, we explore the clinical utility of digital twins in navigating the “de-escalation dilemma” for human papillomavirus–associated disease and improving survivorship through digital phenotyping. With the advent of longitudinal data from medical-grade Internet-of-Things sensors and circulating tumor DNA kinetics, digital twins can serve as active sentinels, detecting functional decline or subclinical events months before clinical manifestation. Furthermore, the implementation of synthetic control arms via virtual populations offers an avenue to accelerate clinical trials and address ethical issues in rare molecular subtypes. Despite this potential, substantial barriers to implementation remain, including profound anatomical instability during treatment, the “black box” nature of deep learning algorithms, and the challenges of multiscale data integration. We provide a technical roadmap for the development of “morphing” digital twins—dynamic systems capable of continuous geometric and biological auditing. By bridging the gap between the macroscopic world of radiation physics and the microscopic world of cellular evolution, digital twins may help shift the management of HNSCC from population-based statistical probability toward more individualized, biologically informed decision-making, with the goal of optimizing HNSCC management while preserving essential patient functions.

Turing · aitrainer.work

Discover how to develop learners’ multimodal literacy skills for today’s digital world. This practical session explores strategies, activities, and techniques for integrating multimedia communication…