最新模型

新模型、開放權重、多模態、語音、影像與模型架構。

LangChain Community Reports Summary Validation Gap That Could Cause Agents to Lose Working Context

A LangChain community report dated October 20 says the middleware may remove original messages when a summarization model returns blank text, silently leaving agents without context. The report includes a minimal reproduction and a proposed fix, but no maintainer confirmation or verifiable release containing a fix is available yet.

2026-10-21

GitHub Repo

AWS Publicly Tests TorchNeuron, Bringing Native PyTorch Execution to Trainium; Compilation Still Limited to Static Shapes

On October 20, AWS released the open beta of open-source TorchNeuron, supporting PyTorch eager execution and model compilation on Trainium2 and Trainium3. The source documentation also lists limitations for dynamic shapes, data types, and operator compatibility, so migrations still require item-by-item validation.

2026-10-21

GitHub Repo

Transformers Community Revisits RT-DETR Training Loss Mismatch; Legacy Denoising Queries May Crowd Out Positives

A report to the Transformers community on October 21 said RT-DETR models trained with version 5.14.1 may mix denoising queries into the main loss, weakening positive supervision for regular queries. A related fix was merged to the main branch in September, but the reported experiment has not been independently reproduced; users should check their installed version.

2026-10-21

GitHub Repo

Open WebUI Community Reports an MCP Authorization Selection Gap: Configured Secrets May Still Not Be Sent as Client Credentials

A report filed with the Open WebUI community on October 21 says static OAuth configuration may fail to authorize MCP tools because of the order in which a server lists supported authentication methods. The report includes reproduction steps and a proposed change, but maintainers have not confirmed it and no official fix has been released.

2026-10-21

GitHub Repo

Transformers report says Gemma 2 attention soft cap is skipped; switching backends may change model computation

A report filed on October 21 says Gemma 2 may silently skip its attention-logit soft cap when using default SDPA or some Flash Attention kernels. It includes numerical comparisons and reproduction code, but no official fix version has been confirmed, and the findings do not establish quality changes across all tasks.

2026-10-21

GitHub Repo

Hermes Agent Community Reports a Gap in Visual History: Gateway Conversations May Lose Tool-Loaded Images

A report from the Hermes Agent community on October 20 says that when a gateway reloads conversation history, native images supplied by tools may be reduced to text placeholders. This could cause a model to retract an earlier, correct visual interpretation. No confirmed official fix release is available yet.

2026-10-21

GitHub Repo

Transformers Maintainers Confirm a Sampling Gap in Serve; Positive Temperature May Still Use Greedy Decoding

A Transformers report on October 19 found that some models may not enable random sampling when given a positive temperature through serve. Maintainers confirmed the issue and said they would rework the fix, but no confirmed release containing a fix was available.

2026-10-20

GitHub Repo

Transformers Reports Qwen3-MoE Loss Scaling Issue That May Amplify Router Auxiliary Weight in Multi-GPU Training

A Transformers report on October 19 says a specific Trainer setup may scale Qwen3-MoE’s router auxiliary loss with the number of data-parallel processes. A project member has responded with an analysis, but the available measurements use a fixed older version; no fix has been merged or released.

2026-10-20

GitHub Repo

LangChain Community Reports Counting Gap: Tool Limit May End Run Early and Allow Model Calls Past Cap

A LangChain community report on October 19 said that the ordering of two built-in middleware components may leave completed model calls uncounted, disabling cumulative conversation limits. Another contributor said they reproduced the issue and prepared a fix, but no maintainer confirmation or official fix release was available.

2026-09-27

GitHub Repo

Dify Community Reports Queued Nodes Starting After Workflow Stop, Potentially Allowing Later Tool Calls

A Dify community report on October 20 says queued nodes may still start after a workflow receives a stop signal, potentially adding model costs or triggering external actions. A proposed fix includes regression tests but has not been merged; the impact on official releases and the cloud service remains unconfirmed.

2026-10-20

MiniMax Releases M3.1 Flash, Targeting Low-Cost Agent Tasks and Million-Token Context

MiniMax launched M3.1 Flash on October 19, tuning sparse inference to improve speed and lower the cost of agent tasks. The model is available through an API, but its weights have yet to be released, and its official performance figures await independent verification.

2026-10-19

Inception Releases Mercury 3, a Diffusion Language Model with One Million Tokens of Context

Inception launched Mercury 3 on October 19, expanding its context window to one million tokens and claiming output speeds of up to 1,068 tokens per second. Application developers have begun integrating it, but the company’s speed and quality comparisons still await independent evaluation.

2026-10-20