Videos
Find videos about LLM models, hardware, runtimes, benchmarks and practical AI workflows.
Enter Pro compares GPT-6 Astra, Opus 5.5 and local Qwen performance when building full web apps through side-by-side testing.
Benchmarks evaluate Gemini 4 Argon, Claude Fable 5.5, GPT-6.1 Sol Ultrafast, and Step 5 against DeepSeek, GLM 5.3, and Qwen 3.8 using WoAI Bench.
Locally installs and tests pg-jev, a PostgreSQL extension powered by TypeSafe's Jev.
Locally installed Humanizer rewrites AI-written drafts to sound human, tested on work emails using Qwen drafts and social posts about getting rich.
Develops and tests a UI framework for Dust to Dominion using Qwen 3.8 Flash Next via Strata within a spec-driven vibe coding experiment.
Runs Shanghai AI Laboratory's Atria Dawn locally using Inferencer v2.4.4 on M3 Ultra and M4 Max Mac Studios, comparing performance against GLM, Kimi K3 and Opus across various tasks.
Locally installed d1-omni-600M undergoes text, vision, and audio tests using an angry customer refund scenario, Leone Partners image, 13 languages FLEURS dataset, happiness clip, and benchmarks.
Evaluates Claude Haiku 5.5 via Claude Code and WoAI Bench, comparing performance and cost against GPT-6 Luna, Grok 4.7, Claude Sonnet 5.5, and Claude Opus 5.5 across coding, game dev, and benchmarks.
Demonstrates using Jev within an AI second brain via OpenRouter to filter emails, guard against prompt injection, rate news, and catch duplicates with confidence scores under a second.
Evaluates Claude Haiku 5.5 against frontier models via browser OS tests, C++ skate games, Blender & Godot pool games, Subway FPS, RuneScape 2007 replicas, and Demo Derby benchmarks.
Tests Claude Haiku 5.5 using bug fix, GoDot, benchmarks, multilingual, scientific reasoning visual, lawn situation, and movement evaluations.
Tests Qwen3.8-27B-Escha-W2-GGUF performance, memory, reasoning, HumanEval Remix, Kanban, Sand Physics, Dungeon Crawler, Blender and Godot.
Tests running the 125B parameter Qwen3.8-Flash-Next using Strata across various VRAM configurations to evaluate its performance on low VRAM setups.
Locally installs and tests EmbeddingGemma 2, an open multimodal embedding model.
Tests Mistral Large 4 across browser OS, C++, Blender, Godot, frontend design, FPS generation, Chrono City, and image-to-game tasks.
Tests Mistral Large 4, also known unofficially as ML4 and officially as le Chonk.
Compares Cyber-Tiel-Coder against Qwen 3.6 35B A3B using timed challenges to evaluate performance among fast local LLMs.
Demonstrates providing BearQ agents with PRDs, business rules, and documentation to improve testing decisions and automatically create test cases via the QA Lead, Explorer, and Tester Agents.
Compares GPT-Next outputs with GPT-6.1 Sol, evaluates Claude Max versus ChatGPT Pro via SemiAnalysis, and covers OpenAI watermarking, Reflection AI Beam, and Claude Code updates.
Locally installs and tests CodeAF, described as a coding harness, using Ollama.
Demonstrates using Claude Opus 5.5 for game modding via pass-through mods, reverse engineering, rebuilding in Rust, and porting mechanics.
Compares Reflection's 501B open-weight Beam and Cohere's North 2 enterprise platform regarding control over AI.
Evaluates Qwen3.8-Flash-Next with Strata on an RTX 4060 Ti 16GB via Pi agent across three coding tasks while measuring generation speed.
Tests Ling 3.1 Flash via browser OS, C++ skate game, Watch website, Blender & Godot pool, Subway FPS, GTA fix, and Street Yeet game tasks.
Demonstrates Strata running Qwen 3.8 Flash on RTX 3060, achieving 900+ tok/s prompt processing via a voice agent placing a grocery order in under two minutes.
Demonstrates building a Python-based MCP harness using docmgr to integrate with OpenCode and TalkWithMe for executing custom search, read, and edit operations.
Tests Swift 1.5 Qwen3.8-27B · GSQ-RCO performance, memory, reasoning, OpenAI Human Eval, Kanban, Sand Physics, Dungeon Crawler, Blender and Godot against its claimed token reduction and speed-up.
Collection of LoRAs for face, head, and body replacement across Qwen Image 2.1, Qwen Image Edit 2509/2511, Flux 2 Klein, Krea 2, and experimental LTX-2 workflows.
Runs Inferencer v2.4.4 with DeepSeek-V4.1-MLX-Q4i and Q9 variants on M3 Ultra 512GB plus M4 Max 128GB cluster to test local performance against Opus 5 coding benchmarks.
Locally installs and tests Decision-2.0-Nox-4B, the 4B model of Decision 2.0, which serves as the decision models of vLLM Semantic Router.
Tests Aleph Alpha's Kolibri, a mixture-of-experts (MoE) reasoning model.
Demonstrates using Claude Opus 5.5 with Higgsfield MCP to autonomously generate the Texodus landing page and an 8-second promo video from minimal prompts.
Locally installs and tests Ubroker, which locates personal info exposure across data brokers and automates removal requests using NousResearch hermes-agent optional skills.
Locally installed Microsoft FrogNano-4B-2609 is tested for its focus on repository-level software engineering tasks.
Evaluates Context Language Models managing context via Suffix Cache Reuse, showing improved accuracy and reduced FLOPs on BrowseComp-Plus and EdgeBench benchmarks.
Tests Qwen3.8 variants on RTX PRO 6000 via SGLang/vLLM across 10 tasks including BFCL, long context, SVG, video editing, voxels, motion design, Rube Goldberg physics and CAPTCHAs.
Locally installed and tested Limite 1B - Violetto, a model designed for high-throughput solutions of difficult mathematical problems.
Covers OpenAI Dots, GPT-6.1 Sol, Claude Sonnet 5.5, Gemini 4 Argon, Ideogram 4.5, Strands Decider 2B, xAI Team Bots, Eleven v4, Microsoft MAI-Voice-2.1, and Trump superintelligence order.
Tests Mac mini M6 behavior under concurrent loads of 8 Camtasia 4K exports, local RAG via Ollama, Gemma 4 inference and 4K playback across CPU, GPU, memory, media engines and storage.
Locally installs and tests Clef, a 27B multimodal model that converts a state and a schema of typed questions into decisions using Cloudflare/clef resources.
Tests MiniMax M3.1 Flash Preview via MiniMax Code on AI coding, games, 3D environments, and websites, comparing speed and output against GPT-6 Sol and Fable 5.1.
Tests local AI on a base M6 Mac mini using Ollama, comparing Gemma 4 26B against Gemma 4 E4B MLX for responsiveness during a setup speed run.
Tests MAI-Voice-2.1 and MAI-Transcribe-2 for accurate, fast, low-cost audio understanding and generation to build conversational voice agents.
Compares GPT-6.1 Sol and Claude Sonnet 5.5 via demolition derby games, robot product design projects, parallel parking simulations, and creative audio-visual tasks.
Compares M5 Ultra against two DGX Sparks, noting the M5 Ultra requires an unspecified factor to surpass the DGX Sparks performance.
ISTA-DASLab demonstrates running QWEN3.8-FLASH-NEXT-GSQ-RCO in LM Studio and compares its performance against QWEN3.8-Q4 for .NET migration and SVG image generation tasks.
Tests K2 Horizon MoVA 36B A4B using Performance, Memory, Reasoning, HumanEval Remix, Kanban, Sand Physics, Dungeon Crawler, Blender and Godot tasks.
Tests Ling 3.1 Flash, InclusionAI's hybrid reasoning language model, for coding, multi-step analysis, and tool-using agents.
Early tests evaluate Gemini 4 Argon's long-horizon reasoning, coding, agentic capabilities, 1 million token output limit, pricing, hallucination rates and performance via WoAI Bench and Arena.