Loading activity…
@artificialanlys
Independent analysis of AI
AI
ArtificialAnlys puts GPT-6 Luna (high) at $0.03 per task versus $2.49 for the first model to reach 30.
AI
ArtificialAnlys says it tested Fibo 2.0 through Bria's API at default settings and found gains over earlier Fibo models.
AI
The SDK streams listening, reasoning, tool calls and spoken replies in one pipeline, Mistral says.
AI
Artificial Analysis put Ultra fifth on AA-Image-Editing v2.1; Arena ranked it fourth in Multi-Image Edit on October 20.
AI
Artificial Analysis ranks it third among open-weights models and 19th overall when models use their native voices.
AI
Artificial Analysis launched both leaderboards; a 2,000-character length adjustment changes some models’ rankings.
AI
Cognition says Devin Fusion beats Claude Code with Opus 5.5 on performance and price, at 31% lower cost per task.
AI
Artificial Analysis points readers to its image-to-video and text-to-video leaderboards and invites votes in its Video Arena.
AI
fal reports a 12.45-second generation time for a 10-second clip in one comparison, not a guaranteed wait.
AI
Fal says the model, developed with @bfl_ai, generates a 10-second 720/768p video in about 12 seconds.
AI
Artificial Analysis uses 900 questions to assess 30 judged traits and 21 separately measured writing characteristics.
AI
Artificial Analysis says Exa's auto mode with Dynamic Highlights scored 80 on its Search Index, tying Perplexity Search (medium).
AI
The launch post claims the diffusion LLM outperforms GPT-6 Luna and Haiku 5.5 on quality.
AI
A post quotes API reference docs describing MiniMax-M3.1-Flash as a frontier multimodal coding model.
AI
A user says the model regresses on knowledge versus its predecessor and on LCR.
Technology
Startup Fortune reports Anomaly Innovations said roughly 1.4% of OpenCode Console users were affected.
AI
DeepSeek says the model is live on its website, app and API, and optimized for DeepSeekHarness and Tencent CodeBuddy.
AI
ArtificialAnlys says users can create personalized benchmarks in three steps and compare models on intelligence, cost and speed.
AI
Reflection says the release includes Mirror CLI, its open-source training harness, and collaborations with 40-plus partners.
AI
Artificial Analysis reports d1-ultra’s non-hallucination rate rose from d1’s 2% to 85%, while accuracy fell from 41% to 32%.
