Models
66 RohitAI articles on Models: news, explainers and practical guides, newest first.
Cognition launches SWE-2 Flash with a free offer through November 30
SWE-2 Flash arrives in Devin Desktop and CLI with free use advertised through November 30, but account eligibility and later prices remain unclear.
Antigravity’s November 2 model deadline changes Claude access
Three Antigravity models are scheduled for removal on November 2. See which Gemini and Claude choices remain, and how paid Pro differs from a trial.
Claude Opus 4.8 fast mode ends November 30
Choose between Opus 4.8 at standard speed and a move to Opus 5.5 before the November 30, 2026 cutoff, with pricing and migration checks.
Transformers adds adapter training for FP8-compressed models
Teams can customize supported FP8-compressed models with adapters in Transformers. Hardware and training-backend limits determine which setups qualify.
Mistral releases Large 4 weights for self-hosting
Mistral’s Large 4 downloads let businesses run and adapt the model themselves. Hardware needs, API prices and licence conditions shape the choice.
Liquid AI’s LFM3 brings pictures, video and speech to local assistants
Liquid AI’s downloadable LFM3-Omni combines visual and spoken inputs. Software support, context limits and commercial-use terms shape adoption.
OpenAI launches lower-priced GPT-6.1 Terra for everyday work
GPT-6.1 Terra brings a lower-priced option to ChatGPT Work, Codex and the API. Here is who can use it and what its lower token rates mean.
Elorian introduces Glimpse for chart and video analysis
Elorian’s new visual AI targets chart reading and video tagging. What its launch evidence, token prices and access limits mean for teams considering it.
Reflection releases Beam’s weights for custom coding assistants
Beam’s downloadable weights let businesses customize and run the coding model themselves. Its 273 GiB compressed download still needs substantial server capacity.
What NVIDIA Nemotron Bolt’s 24GB target means for local AI
NVIDIA’s Bolt model is available for local AI assistants. Its 24GB text-serving target excludes vision; the deployment recipes clarify what hardware to plan for.
NVIDIA releases Nemotron Super VL for custom document assistants
NVIDIA’s new model can be adapted for business documents. Its training requirements and free-trial data limits matter before commissioning an assistant.
OpenAI sets January 14, 2027 shutdown for selected GPT-4 API models
Five GPT-4 API versions and their fine-tunes will shut down on January 14, 2027. Businesses need to check customization and tool compatibility before moving.