EVENT DATE
GLM-5.3-Flash technical overview details native vision
AutoClaw describes GLM-5.3-Flash's multimodal architecture. The model has 320B total parameters and activates 18B.
FINDGOODAI / MODEL RADAR
Model Radar
Track model releases, capabilities, pricing, open weights and API changes.
Explore the latestRELEASES · CAPABILITIES · PRICING · OPEN WEIGHTS · APIs
THE UPDATE LOG
Latest updates
17 updates
EVENT DATE
Gemini 4 Argon starts with trusted cyber defenders
Google announced Gemini 4 Argon, initially accessible through Fairwind. Wider access for developers and consumers remains a later step.
Source Google (opens in a new tab)
EVENT DATE
GPT-6.1 Sol: Access, Workflow Diagrams, and a Guide to Complex Work
Understand access and API settings, explore practical coding and document workflows, and use two official diagrams to explain multi-agent execution and task cost.
Source OpenAI (opens in a new tab)
EVENT DATE
Claude Sonnet 5.5 targets lower task costs at unchanged rates
Anthropic reports over 30% faster output and task savings of up to 30% versus Sonnet 5, with token rates unchanged.
EVENT DATE
Claude Opus 5.5 cuts input and output rates by 20%
Anthropic released Opus 5.5 with lower token rates. Its claimed 40% task-cost reduction measures something different.
EVENT DATE
Grok 4.7 upgrades coding at unchanged base rates
Grok 4.7 targets longer coding and knowledge-work tasks. The announcement retains Grok 4.6's base pricing and serving speed.
EVENT DATE
Qwen-Image-2.1 unifies editing and transparent image generation
Qwen announced an open-source image model combining generation and editing. Its visual generation component has seven billion parameters.
Source Qwen (opens in a new tab)
EVENT DATE
Qwen3.8-LiveTranslate adds speaker-aware bilingual output
Qwen's interpretation update adds speaker separation and synchronized source-and-translation output. Developer testing reports average lag falling from 2.8 to 2.3 seconds.
Source Qwen (opens in a new tab)
EVENT DATE
Grok Voice Transcribe 2.0 keeps $0.10 hourly batch pricing
Grok's new transcription model targets multilingual and difficult audio. Batch and streaming access retain previous rates.
EVENT DATE
Jev turns AI judgment into a function: charts, Playground and Python integration
Six launch and cookbook figures explain typed decisions, cost comparisons, re-ranking and a practical Python integration for TypeSafe’s Jev.
EVENT DATE
Kimi K2.8 Preview reaches all Kimi Code users
Kimi completed the K2.8 Preview rollout in Kimi Code. Existing clients retain the kimi-for-coding identifier.
EVENT DATE
GPT-Live-1 reaches the API at $0.05 per voice minute
OpenAI's full-duplex model can listen while speaking. It delegates deeper reasoning and tool use to backend models.
Source OpenAI (opens in a new tab)
EVENT DATE
DeepSeek V4.1 Flash adds vision and redirects older APIs
DeepSeek released a native multimodal model under deepseek-flash. Its architecture separates input and output computation.
EVENT DATE
GPT-6 Astra explained: long workflows, tools, API usage and task cost
A practical look at Astra’s context, async tools, mid-turn steering and Ultrafast, with concrete workflows, API examples and a method for evaluating completed work.
Source OpenAI (opens in a new tab)
EVENT DATE
Gemini 3.8 Flash arrives alongside a restricted Cyber variant
Google introduced Flash for agent workflows and Flash Cyber for defense. Flash retains 3.7 Flash's introductory rates.
Source Google (opens in a new tab)
EVENT DATE
Fable 5.1 and Mythos 5.1 launch with different access rules
Anthropic says the releases share a model but use different safeguards. Fable is generally available; Mythos requires trusted access.
EVENT DATE
GLM-5.3 advances coding and requires thinking-enabled requests
Z.ai released GLM-5.3 using the GLM-5.2 base with expanded post-training. Disabling thinking is no longer supported.
Source Z.ai (opens in a new tab)