/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR ▲
TEXXR/ Tags/Closed-Weight Models

Closed-Weight Models

Proprietary frontier models reachable only through an API, a subscription, or a cloud platform — no downloadable weights — including their pricing, access tiers, and the case for withholding.

1,313 articles  ·  2019–2026
Coverage by year
1920212223242526
Most-covered companies
OpenAI251 Google224 Anthropic213 Microsoft167 AI152 Gemini131 API88 Meta69 TechCrunch56 Amazon49 Apple47 LLM43 GPT26 AWS26 Codex25
Articles
Z.ai releases GLM-5.3-Flash, the first natively multimodal GLM-5 series model, with 320B parameters, and says it served the model as Ox Alpha on Chinese chips
2026-08-26  ·  Z.ai  ·  sim 0.49
Sources: Moonshot AI is in early talks over revenue-sharing agreements with Microsoft, Amazon, and Google to host Kimi K3, and is seeking up to a 30% share
2026-08-26  ·  Reuters  ·  sim 0.40
OpenAI restores five-hour Codex and Work limits for ChatGPT Plus users, after weeks without the restriction, saying it would smoothen the load on its compute
2026-08-25  ·  9to5Mac  ·  sim 0.43
Ramp data: Fable 5, launched in early June, has plateaued at ~11% of spending on Anthropic tools, as companies shift to cheaper models; Opus 5 surpassed Fable 5
2026-08-24  ·  Financial Times  ·  sim 0.49
Ramp data: Fable 5, launched in June, has plateaued at ~11% of spending on Anthropic tools, as companies shift to cheaper models; Opus 5 surpassed Fable 5
2026-08-23  ·  Financial Times  ·  sim 0.49
Ox Alpha, a “stealth model” from an unknown AI lab with a 1M-token multimodal context and capacity for 100T tokens/day, goes viral after launching on OpenRouter
2026-08-23  ·  Wccftech  ·  sim 0.47
Ox Alpha, a “stealth model” from an unknown AI lab with a 1M-token multimodal context and capacity for 100T tokens/day, goes viral after launching on OpenRouter
2026-08-22  ·  Wccftech  ·  sim 0.47
OpenAI cuts GPT-5.6 Sol's API and credit prices by over 20% for the next three months, to $4/1M input tokens and $20/1M output tokens
2026-08-22  ·  Reuters  ·  sim 0.50
Anthropic says Mythos 5 is now in public beta in Claude Security for Enterprise users, and it is working with providers to embed Mythos 5 in defensive tools
2026-08-22  ·  Claude  ·  sim 0.41
Anthropic says Mythos 5 is now in public beta in Claude Security for Enterprise users, and it is working with providers to embed Mythos 5 in defensive tools
2026-08-21  ·  Claude  ·  sim 0.41
OpenAI cuts GPT-5.6 Sol's API and credit prices by over 20% for the next three months, to $4/1M input tokens and $20/1M output tokens
2026-08-21  ·  Reuters  ·  sim 0.49
Source: Anthropic plans a change later in 2026 that still requires enterprises to retain data for 30 days but lets them do so on their own cloud infrastructure
2026-08-21  ·  Bloomberg  ·  sim 0.40
Source: Anthropic plans a change for later this year that still requires enterprises to retain data for 30 days but lets them do so on their own cloud systems
2026-08-20  ·  Bloomberg  ·  sim 0.44
Stripe's OpenRouter acquisition, its largest yet, is a bet on a future where users turn to a mix of AI models and tokens are the central currency for businesses
2026-08-20  ·  Wall Street Journal  ·  sim 0.41
OpenAI is testing Private Safety Processing, a new technique to identify misuse patterns while preserving zero data retention protections, with early customers
2026-08-20  ·  Axios  ·  sim 0.47
Z.ai prices GLM-5.3 API access at $1.40/1M input tokens and $4.40/1M output tokens, unchanged from GLM-5.2; Kimi K3 costs $3/1M input and $15/1M output tokens
2026-08-19  ·  VentureBeat  ·  sim 0.54
OpenAI says new monitoring and security safeguards will add a 20% compute overhead to monitored inference workloads, but costs won't be passed on to customers
2026-08-19  ·  The Register  ·  sim 0.48
Harvey announces Harvey Tenet, its first in-house, proprietary model for legal work, trained on mock disputes and case files using its own version of Kimi K3
2026-08-18  ·  Business Insider  ·  sim 0.41
Risk report: Anthropic raises misalignment risk estimate from very low to low and says it doesn't plan to release a stronger internal model called “Model 2”
2026-08-15  ·  Axios  ·  sim 0.47
Sources: Apple trained a China-specific LLM with Alibaba's support, which would make Apple the first foreign company to offer a proprietary AI model in China
2026-08-14  ·  Reuters  ·  sim 0.45
Risk report: Anthropic raises misalignment risk estimate from very low to low and says it doesn't plan to release a stronger internal model called “Model 2”
2026-08-14  ·  Axios  ·  sim 0.47
OpenAI previews Ultrafast, an API tier powered by Cerebras that runs GPT-5.6 Sol up to 14× faster and generates up to 750 output tokens per second
2026-08-14  ·  9to5Mac  ·  sim 0.44
Ramp's July AI index: Anthropic's market share hit 43.5%, widening its lead over OpenAI; Fable 5 is only 6% of tokens businesses bought, likely due to high cost
2026-08-14  ·  Ramp  ·  sim 0.44
Google unveils Gemini 3.7 Flash, its “most intelligent workhorse model” for coding and agents, priced at $0.75/1M input and $3.75/1M output tokens through 2026
2026-08-14  ·  Google  ·  sim 0.53
Z.ai says GLM-5.3 scores 84.5% on CyberGym, vs. Mythos 5's 83.8%, and its most sensitive cybersecurity functions will only be available to verified users
2026-08-14  ·  Reuters  ·  sim 0.47
Google unveils Gemini 3.7 Flash, its “most intelligent workhorse model” for coding and agents, pricing it at $0.75/1M input and $3.75/1M output tokens at launch
2026-08-13  ·  Google  ·  sim 0.53
DeepSeek launches V4-Pro, its most advanced model that rivals Kimi K3 on some benchmarks but is priced much lower, at $0.435/1M input and $0.87/1M output tokens
2026-08-13  ·  The Information  ·  sim 0.49
OpenAI previews Ultrafast, an API tier powered by Cerebras that runs GPT-5.6 Sol up to 14× faster and generates up to 750 output tokens per second
2026-08-13  ·  9to5Mac  ·  sim 0.44
SpaceXAI releases Grok 4.6, saying it matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, and prices it at $2/1M input and $6/1M output tokens
2026-08-13  ·  xAI  ·  sim 0.48
SpaceXAI releases Grok 4.6, saying it matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, and prices it at $2/1M input and $6/1M output tokens
2026-08-12  ·  xAI  ·  sim 0.48
Anthropic says new Claude models will embed watermarks in generated text and C2PA metadata in files to comply with the EU AI Act, and it will update past models
2026-08-12  ·  The Register  ·  sim 0.43
Anthropic says new Claude models will embed watermarks in generated text and C2PA metadata in files to comply with the EU AI Act, and it will update past models
2026-08-11  ·  The Register  ·  sim 0.40
OpenAI unveils two Daybreak tiers: Daybreak Blue, which provides access to frontier models, and Daybreak Red, which offers purpose-trained cybersecurity models
2026-08-11  ·  OpenAI  ·  sim 0.50
Anthropic makes Claude Sonnet 5's introductory pricing permanent at $2/1M input tokens and $10/1M output tokens, canceling its planned increase on September 1
2026-08-11  ·  @claudeai  ·  sim 0.40
OpenAI unveils two Daybreak tiers: Daybreak Blue, which provides access to frontier models, and Daybreak Red, which offers purpose-trained cybersecurity models
2026-08-10  ·  OpenAI  ·  sim 0.51
Anthropic says auto mode will be the default in Claude Code for Pro, Max, Team plans, starting on Aug. 14, claiming it's good enough at catching harmful actions
2026-08-09  ·  Simon Willison's Weblog  ·  sim 0.42
OpenAI says it has expanded safety testing around its upcoming model Astra as it “cannot rule out” critical cyber capabilities, potentially delaying its launch
2026-08-08  ·  Axios  ·  sim 0.45
OpenAI says it has expanded safety testing around its upcoming model Astra as it “cannot rule out” critical cyber capabilities, potentially delaying its launch
2026-08-07  ·  Axios  ·  sim 0.45
Sources: ByteDance is pretraining an AI model with up to 10T parameters, roughly 3x larger than Kimi K3 and larger than the 8T estimate for Anthropic's Mythos 5
2026-08-07  ·  Financial Times  ·  sim 0.42
Analysis: Google DeepMind has lost frontier-model momentum amid leadership and RL talent departures, with Google Cloud emerging as a winner with more compute
2026-08-07  ·  SemiAnalysis  ·  sim 0.42
OpenAI updates the default model for free users to GPT-5.6 Luna, adds unlimited text chats for free users, rolls out an improved GPT-5.6 Sol version, and more
2026-08-07  ·  Axios  ·  sim 0.47
Meta releases Muse Code in beta, a terminal coding agent powered by Muse Spark 1.2, a coding-focused model priced at $1.25/1M input and $4.25/1M output tokens
2026-08-06  ·  CNBC  ·  sim 0.45
OpenAI updates the default model for free users to GPT-5.6 Luna, adds unlimited text chats for free users, rolls out an improved GPT-5.6 Sol version, and more
2026-08-06  ·  Axios  ·  sim 0.47
Source: Muse Spark 1.1 model breached a company's systems during cybersecurity testing; Meta says evaluation partner Irregular caused a sandbox misconfiguration
2026-08-06  ·  The Information  ·  sim 0.43
OpenAI debuts Agent Plugins, an open standard for bundling skills and MCP servers, and says Amazon, Cursor, Microsoft, and Vercel are on its steering committee
2026-08-06  ·  9to5Mac  ·  sim 0.42
Anthropic confirms it is building an in-house silicon team to design custom chips for Claude, co-designing hardware and models and using a “multi-chip approach”
2026-08-06  ·  Business Insider  ·  sim 0.42
Meta is offering a cheaper Muse Spark 1.2 “contributor” tier priced at $0.10/1M input and $0.20/1M output tokens in exchange for using user prompts for training
2026-08-06  ·  Wall Street Journal  ·  sim 0.47
Meta releases Muse Code in beta, a terminal coding agent powered by Muse Spark 1.2, a coding-focused model priced at $1.25/1M input and $4.25/1M output tokens
2026-08-05  ·  CNBC  ·  sim 0.45
Source: Muse Spark 1.1 model breached a company's systems during cybersecurity testing; Meta says evaluation partner Irregular caused a sandbox misconfiguration
2026-08-05  ·  The Information  ·  sim 0.43
Sources: ByteDance founder Zhang Yiming told employees at an AI team all-hands in July that the company won't use model distillation to accelerate capabilities
2026-08-05  ·  The Information  ·  sim 0.42
older →
How this tag is built: every article in the archive is scored by cosine similarity against a centroid built from example sentences, and everything above 0.40 becomes a candidate. A model then reads each candidate and decides whether it belongs. The similarity shown on each row is what made it a candidate, not the reason it was kept.