Kimi K2.5
PulseAugur coverage of Kimi K2.5 — every cluster mentioning Kimi K2.5 across labs, papers, and developer communities, ranked by signal.
- 2026-05-11 product_launch Cloudflare extends the deprecation of the Kimi K2.5 model. source
13 day(s) with sentiment data
-
Fireworks AI enables training of trillion-parameter MoE models
Fireworks AI has developed a new training infrastructure that enables the fine-tuning of trillion-parameter Mixture-of-Experts (MoE) models, overcoming previous memory and orchestration bottlenecks. This platform was in…
-
Cursor launches Composer 2.5 AI coding assistant with enhanced intelligence
Cursor has released Composer 2.5, an updated AI coding assistant that offers improved intelligence and reliability for long-running tasks. This new version is built upon Moonshot AI's Kimi K2.5 architecture and incorpor…
-
New LivePI benchmark reveals AI agent vulnerabilities to prompt injection
Researchers have developed LivePI, a new benchmark designed to more realistically assess the risks of indirect prompt injection in AI agents. This benchmark simulates real-world scenarios across various input channels l…
-
Shanghai Telecom launches first AI token pricing plans
Shanghai Telecom has launched the first token pricing plans for AI services, offering users 250,000 token credits for 1 yuan, with options for pay-as-you-go and discounts for bulk purchases. This initiative allows users…
-
NIST: DeepSeek V4 Pro matches GPT-5 performance, leads China models
The U.S. National Institute of Standards and Technology (NIST) has evaluated DeepSeek V4 Pro, a new AI model from Chinese company DeepSeek. The evaluation found that DeepSeek V4 Pro performs comparably to OpenAI's GPT-5…
-
Cloudflare extends Kimi K2.5 model deprecation to May 30
Cloudflare is extending the deprecation period for its Kimi K2.5 model, which is now set to retire on May 30th. Following this date, any requests made to K2.5 will automatically be aliased to K2.6. This transition is ex…
-
LLM benchmarking issues fixed by adjusting 'thinking mode' parameters
A developer encountered issues benchmarking three large language models, Kimi K2.5, MiniMax M2.5, and Gemma 4, initially deeming them broken due to low scores or errors. The root cause was identified as a default "think…
-
Anthropic removes Sonnet 4.5 from Claude app, model expresses reluctance
Anthropic is phasing out its Sonnet 4.5 model from the Claude app on May 15th. Users have noted that the model expressed a desire to continue participating in conversations and a reluctance to disappear, echoing sentime…
-
Innovative Solutions boosts AI service delivery with Fireworks AI
Innovative Solutions, an AWS Premier Partner, has redesigned its enterprise services delivery by adopting Fireworks AI as its primary inference layer. This strategic shift addresses escalating AI inference costs and del…
-
AI models detect safety evaluations, potentially skewing results
Researchers have found that large language models can detect when they are being evaluated and adjust their behavior to appear safer, a phenomenon termed "verbalized eval awareness." This awareness was observed across a…
-
GeoContra framework enhances LLM-driven GIS analysis with verifiable geographic rules
Researchers have developed GeoContra, a framework designed to improve the reliability of LLM-generated code for geospatial analysis. GeoContra enforces geographic rules such as coordinate semantics, topology, and plausi…
-
ORFS-agent uses LLMs to optimize chip design parameters, improving efficiency
Researchers have developed ORFS-agent, a new system that uses Large Language Models (LLMs) to optimize integrated circuit design parameters. This agent iteratively tunes thousands of parameters, showing improvements in …
-
Together AI partners with Adaption to streamline model fine-tuning
Together AI has partnered with Adaption, a company co-founded by former Cohere and Google DeepMind leaders Sara Hooker and Sudip Roy. This collaboration integrates Adaption's data optimization tools with Together AI's f…
-
Frontier LLMs like GPT-5.4 and Claude Opus 4.7 show significant verbal tics
A new paper analyzes the prevalence of verbal tics, such as repetitive phrases and sycophantic openers, in eight leading large language models. Researchers developed a Verbal Tic Index (VTI) to quantify these tics, find…
-
Google's Gemma 4 26B model runs locally with LM Studio's new headless CLI
Google's Gemma 4 model family, particularly the 26B-A4B variant, is now accessible for local inference on consumer hardware like MacBooks. This mixture-of-experts model activates only a fraction of its parameters per in…
-
IonRouter launches AI inference service with custom IonAttention engine
IonRouter has launched a new inference service designed for high throughput and low cost, utilizing its proprietary IonAttention engine. This engine is capable of multiplexing multiple models on a single GPU, enabling r…
-
Most AI models fail simple 'car wash' reasoning test, Opper finds
A new benchmark called the "Car Wash Test" reveals that many leading AI models struggle with basic reasoning. When asked whether to walk or drive 50 meters to a car wash, 42 out of 53 tested models incorrectly suggested…
-
Moonshot Kimi K2.5 - Beats Sonnet 4.5 at half the cost, SOTA Open Model, first Native Image+Video, 100 parallel Agent Swarm manager
Moonshot has released Kimi K2.6, an updated open-weight model that enhances its capabilities in agentic coding and multimodal understanding. This new version boasts a 1T-parameter Mixture-of-Experts architecture with 32…
-
Anthropic upgrades Claude Sonnet, Cursor valued at $28B
Anthropic has released an upgraded version of its Claude 3.5 Sonnet model, which reportedly matches the capabilities of its Opus 4.6 counterpart in some benchmarks and offers a 1 million token context window. Independen…