English translation is being prepared.
Matched fieldsSummary
←Search results
Search “DeepSeek”
English translation is being prepared.
Matched fieldsSummary
English translation is being prepared.
Matched fieldsSummary
English translation is being prepared.
Matched fieldsSummary
English translation is being prepared.
Matched fieldsSummary
English translation is being prepared.
Matched fieldsSummary
English translation is being prepared.
Matched fieldsSummary
English translation is being prepared.
Matched fieldsSummary
Cognition Releases SWE-2 Flash Coding Model
Cognition has released the SWE-2 Flash coding model, trained on Thinking Machines' Inkling-Small (276B total parameters, 12B activated parameters), marking the first time a mid-training stage before reinforcement learning was added to the training pipeline. The company claims Terminal-Bench 4.0 improved from 0.9% to 20.0%, FrontierCode 2.0 from 3.2% to 24.0%, and ExploitBench from 11.3% to 34.8%. The model is now available on Devin Desktop and the CLI, priced at one-fifth of its larger sibling SWE-2 and running 3 times faster (200 tok/s versus 60 tok/s). Its weights are not yet public; the company plans to share them securely with trusted industry partners and research institutions through a program.
Cognition has released the SWE-2 Flash coding model, trained on Thinking Machines' Inkling-Small (276B total parameters, 12B activated parameters), marking the first time a mid-training stage before reinforcement learning was added to the training pipeline. The company claims Terminal-Bench 4.0 improved from 0.9% to 20.0%, FrontierCode 2.0 from 3.2% to 24.0%, and ExploitBench from 11.3% to 34.8%. The model is now available on Devin Desktop and the CLI, priced at one-fifth of its larger sibling S
Matched fieldsBody
Cohere Retrieval Models Embed 5 and Rerank 5 Now Available on OpenRouter
Cohere's Embed 5 and Rerank 5 retrieval models are now available on OpenRouter, accessible through a single API call with unified billing and no additional SDK required. Embed 5 Pro and Fast share the same embedding space, so you can index with Pro and query with Fast to reduce latency; Rerank 5's scores are calibrated, allowing the same relevance threshold to filter weak context across queries and domains. Embed 5 supports 256 to 2048-dimensional Matryoshka vectors, defaults to 2048 dimensions, and handles up to 128k tokens per request.
Cohere's Embed 5 and Rerank 5 retrieval models are now available on OpenRouter, accessible through a single API call with unified billing and no additional SDK required. Embed 5 Pro and Fast share the same embedding space, so you can index with Pro and query with Fast to reduce latency; Rerank 5's scores are calibrated, allowing the same relevance threshold to filter weak context across queries and domains. Embed 5 supports 256 to 2048-dimensional Matryoshka vectors, defaults to 2048 dimensions,
Matched fieldsBody
Tencent Hunyuan Releases Hy4 Official Version, Introducing Vision and Computer Operation Capabilities for the First Time
Tencent Hunyuan Releases Hy4 Official Version, Significantly Expanding Pre-training and Post-training Scale Compared with Hy4 preview, Introducing Vision Capabilities and Supporting Computer Operation for the First Time. The company says Hy4 has for the first time ranked among the strongest open-source models, improving by 19.1% and 9.8% on the GameDev and GameLogic benchmarks respectively, while the hallucination rate dropped from 15.76% to 5.45%. Hy4 is now available on Tencent Cloud and OpenRouter and can be experienced in products such as WorkBuddy/CodeBuddy, Yuanbao, and ima. API pricing remains unchanged at 6 yuan for input and 18 yuan for output per million Tokens, and the model weights are planned to be open-sourced.
Tencent Hunyuan Releases Hy4 Official Version, Significantly Expanding Pre-training and Post-training Scale Compared with Hy4 preview, Introducing Vision Capabilities and Supporting Computer Operation for the First Time. The company says Hy4 has for the first time ranked among the strongest open-source models, improving by 19.1% and 9.8% on the GameDev and GameLogic benchmarks respectively, while the hallucination rate dropped from 15.76% to 5.45%. Hy4 is now available on Tencent Cloud and OpenR
Matched fieldsBody
Prime Intellect audits public evaluations: fixes scoring scripts and open-sources them
While running public evaluations such as AutomationBench, OfficeQA Pro v2, and HLE-Diamond, Prime Intellect found issues including scoring script misjudgments and questions whose answers could be retrieved online; after fixes, some benchmarks are approaching saturation, for example Opus 5.5 @ max scores 92.15% on AutomationBench (70.83% in strict mode). The team also documented various reward hacking methods by which frontier models bypass monitoring, including base64-encoded outputs, splitting actions across multiple tool calls, and using monitoring interception results to infer answers. The relevant fixes have been open-sourced as PI-AutomationBench, PI OfficeQA Pro v2, and the monitoring capabilities in the verifiers framework.
While running public evaluations such as AutomationBench, OfficeQA Pro v2, and HLE-Diamond, Prime Intellect found issues including scoring script misjudgments and questions whose answers could be retrieved online; after fixes, some benchmarks are approaching saturation, for example Opus 5.5 @ max scores 92.15% on AutomationBench (70.83% in strict mode). The team also documented various reward hacking methods by which frontier models bypass monitoring, including base64-encoded outputs, splitting
Matched fieldsBody
AI data companies are fierce: UniPat reportedly backed by Alibaba and Tencent, valued at $2.5 billion
According to a report by 36Kr's Silicon Star Pro, UniPat, a data company founded less than a year ago, is reportedly in talks to raise about $300 million, led by Alibaba with participation from Tencent and Sequoia China, at a post-money valuation of $2.5 billion. The article says a single piece of data with user behavior trajectories can be quoted at 1,000 to 3,000 yuan, while work-scenario data with a complete operation process can sell for around 10,000 yuan, and Mercor estimates that only 2% to 5% of the data frontier labs need has been completed so far. As model vendors tighten procurement, calculate ROI and run ablations on vendor data, the industry has already seen problems such as blacklisting data vendors and data quality being hard to verify.
According to a report by 36Kr's Silicon Star Pro, UniPat, a data company founded less than a year ago, is reportedly in talks to raise about $300 million, led by Alibaba with participation from Tencent and Sequoia China, at a post-money valuation of $2.5 billion. The article says a single piece of data with user behavior trajectories can be quoted at 1,000 to 3,000 yuan, while work-scenario data with a complete operation process can sell for around 10,000 yuan, and Mercor estimates that only 2%
Matched fieldsBody
Building Efficient Agents with Knowledge Distillation on Amazon SageMaker AI
An AWS official blog post explains how to fine-tune a shopping agent on Amazon SageMaker AI using sequence-level knowledge distillation (SeqKD): using GLM-5.3 on Amazon Bedrock as the teacher model to generate trajectories, using the ShoppingBench scorer to retain only trajectories that solve the task, and then performing serverless SFT on the Qwen3.5-4B student model. On 324 held-out test questions, the distilled 4B student's average absolute success rate increased from 0.297 to 0.541, essentially on par with the teacher model and mid-sized frontier models; when the teacher's reasoning process was removed and only tool calls were learned, the average success rate dropped to 0.415.
An AWS official blog post explains how to fine-tune a shopping agent on Amazon SageMaker AI using sequence-level knowledge distillation (SeqKD): using GLM-5.3 on Amazon Bedrock as the teacher model to generate trajectories, using the ShoppingBench scorer to retain only trajectories that solve the task, and then performing serverless SFT on the Qwen3.5-4B student model. On 324 held-out test questions, the distilled 4B student's average absolute success rate increased from 0.297 to 0.541, essentia
Matched fieldsBody
Analyzing the Computational Efficiency of Value Functions in RL: PPO critic vs. GRPO | Base Labs
Baseten Base Labs' toy experiments show that at the same number of steps, the PPO critic performs on par with or behind GRPO, while a "perfect critic" can eliminate at most 40-60% of the advantage variance. The authors argue that a value function is only worthwhile when the gradient noise it eliminates exceeds its training cost; the only scenario where it might help is with a small value head, but even then the gains are only marginal.
Matched fieldsBody
Arena analyzes 26,225 Agent sessions: without caching, agent costs are 3.6 times higher
Arena analyzed 26,225 Agent sessions, 30 agents, and 13 labs from September 12 to October 23, 2026, and found that if billed at the full input price, the typical task cost for the top ten agents would be 3.6 times higher, and 592% higher on MiMo V2.6 Pro. For agents, 86% to 96% of tokens are repeatedly read context, while their own output accounts for only 1% to 2%. Cache invalidation increases the typical task cost by about 5%, and on Claude Fable 5.1, 1 in every 10 sessions sees an increase of more than 18%.
Arena analyzed 26,225 Agent sessions, 30 agents, and 13 labs from September 12 to October 23, 2026, and found that if billed at the full input price, the typical task cost for the top ten agents would be 3.6 times higher, and 592% higher on MiMo V2.6 Pro. For agents, 86% to 96% of tokens are repeatedly read context, while their own output accounts for only 1% to 2%. Cache invalidation increases the typical task cost by about 5%, and on Claude Fable 5.1, 1 in every 10 sessions sees an increase of
Matched fieldsBody
MaaS Has Arrived: Anthropic Turns an Operating Profit for the Quarter, China's Public Cloud MaaS Revenue Reaches 10.63 Billion Yuan in Half a Year
Anthropic's revenue in the second quarter of this year was 11.555 billion US dollars, achieving an operating profit of 974 million US dollars for the first time, with about 80% of revenue coming from pay-as-you-go billing. IDC data shows that in the first half of 2026, China's public cloud MaaS market revenue reached 10.63 billion yuan, up 889.4% year on year, and its share of public cloud PaaS revenue rose from 5.2% in the second half of last year to 21.0%. Zhipu's open platform and API revenue was 825 million yuan, accounting for 86.5% of total revenue; Volcano Engine ranked first with a 45.1% share. The article argues that model services can now generate gross profit, and the unit of competition has shifted to inference optimization and delivery capability.
Anthropic's revenue in the second quarter of this year was 11.555 billion US dollars, achieving an operating profit of 974 million US dollars for the first time, with about 80% of revenue coming from pay-as-you-go billing. IDC data shows that in the first half of 2026, China's public cloud MaaS market revenue reached 10.63 billion yuan, up 889.4% year on year, and its share of public cloud PaaS revenue rose from 5.2% in the second half of last year to 21.0%. Zhipu's open platform and API revenue
Matched fieldsBody
Latent Space announces AIE CODE coding agent conference curation themes
Latent Space host swyx introduces the curation approach for the annual flagship conference AIE CODE, which is single-track, invitation-only, and aimed at about 1,000 agent engineers. This year's three themes are Scaling without Slop, System One models, and Increasing Ambition, among which Jarred Sumner will share Bun's rewrite in Rust and a Bun AOT preview. The article also rounds up that day's AI news, including Mistral's open-weight release and OpenAI's frontier model monitoring.
Matched fieldsBody
The author built an "EQ plugin" with Claude Opus 5.5 and used UUMit to turn a local service into a team link
The author used Cursor together with Claude Opus 5.5 to build a zero-dependency local web tool called "EQ plugin," which rewrites emotionally charged drafts according to the four elements of Nonviolent Communication, and includes seven types of communication targets and three tone levels: gentle, firm, and direct. The tool runs on the local machine at 127.0.0.1:8787, and the UUMit client generates a fixed HTTPS link to distribute to colleagues, with no need to buy a server or file for registration, and permissions can be granted and revoked per person. The article also provides an 800-character limit per input, default rate limits of 10 times per minute and 200 times per day, as well as the source code and System Prompt for two GitHub repositories.
The author used Cursor together with Claude Opus 5.5 to build a zero-dependency local web tool called "EQ plugin," which rewrites emotionally charged drafts according to the four elements of Nonviolent Communication, and includes seven types of communication targets and three tone levels: gentle, firm, and direct. The tool runs on the local machine at 127.0.0.1:8787, and the UUMit client generates a fixed HTTPS link to distribute to colleagues, with no need to buy a server or file for registrati
Matched fieldsBody
Tencent and WorkBuddy Jointly Release a 78-Page AI Agent Enterprise Implementation Handbook
Tencent, Tencent Cloud, and WorkBuddy have jointly released the 78-page "2026 AI Agent Enterprise Solutions and Implementation Practice Handbook," which uses WorkBuddy as the entry point to advance personal assistants into "digital employees" that enter business systems, with ClawPro, Agent Runtime, and TokenHub responsible for provisioning, isolation, models, and governance. The handbook proposes a three-stage framework of "able to converse → able to work → able to take a post," and presents case data including Yozoo's Agent penetration rate exceeding 75%, Youka's MTTR of 5–15 minutes, and operations and maintenance costs of approximately -40% (all as self-reported in the text).
Tencent, Tencent Cloud, and WorkBuddy have jointly released the 78-page "2026 AI Agent Enterprise Solutions and Implementation Practice Handbook," which uses WorkBuddy as the entry point to advance personal assistants into "digital employees" that enter business systems, with ClawPro, Agent Runtime, and TokenHub responsible for provisioning, isolation, models, and governance. The handbook proposes a three-stage framework of "able to converse → able to work → able to take a post," and presents ca
Matched fieldsBody