$ grep -r "performance" ./posts/

# performance

Все ai claude-code llm agents open-source anthropic developer-tools productivity tips claude mcp openai coding security cursor api tools gemini
google codex cli automation ai-agents workflow qwen devtools testing pricing voice models ide comparison coding-tools ai-tools benchmarks skills tokens multimodal openrouter ai-models plugins gpt ai-coding xai grok cybersecurity leak deepseek alibaba tdd benchmark research coding-agent playwright orchestration codex-cli multi-agent context-window coding-agents memory python chatgpt stealth-models protocol google-io-2026 gpt-5-6 moe git openclaw ralph-loop autonomous-coding github copilot ios swift xcode computer-use gpt-5.4 macos code-review browser-automation unity game-development context-engineering vibe-coding web-scraping kimi browser gemma china tts microsoft cost-optimization glm hunter-alpha video-generation owl-alpha nvidia vision ollama gpt-5.6 local local-llm ai-safety fable meta prompt-engineering safety prompt-injection open-weights coding-assistant deep-research terminal qa php laravel assistant worktrees docker parallel-development oauth websocket context-management mobile perplexity multi-model image-generation remotion video shorts instagram tiktok permissions future code-intelligence knowledge-graph future-of-programming opinion hooks xctest commands local-ai liquid-ai privacy fast-mode copilot-cli linux windows machine-learning cron scheduled-tasks effort settings godot unreal-engine search-api tavily exa agent-teams opus-4.6 expo cowork remote-control plugin google-colab responsive-design frontend telegram discord channels astral superapp licensing documentation prompts figma design web-development demo gamedev gemini-cli speech scraping self-improvement ultraplan debugging function-calling free-tools elevenlabs infrastructure configuration skill dotnet nous-research gpt-6 llama healer-alpha elephant-alpha gpt-5-5 tmux stealth-launch fal elixir linear rust tencent voice-cloning reasoning nemotron mythos policy dense-model game-dev open-beta sonnet gpt-55 spacex managed-agents realtime subq subquadratic long-context transformers finance edge-ai rag vector-search notion typescript workers malware chrome leaks veo lmarena fingerprinting api-pricing onboarding opus-4-8 robotics world-models physical-ai minimax free-models ocr baidu document-ai release-tracker gemini-35-pro ml amazon data-labeling amd hardware llama-cpp code-quality interpretability apple lawsuit curl writing pentesting career wordpress vulnerability evaluation context7 redis devops rce migration alignment kubernetes npm supply-chain cve sqlite dataviz inference vllm performance enterprise
toktier-tokenization-ttft.md
Токенизация съедает до 64% времени до первого токена. Не инференс, а токенизация
> · 8 мин

Токенизация съедает до 64% времени до первого токена. Не инференс, а токенизация

У кодинг-агентов промпт-кеш попадает в 94,1% случаев, и на этом фоне токенизация вырастает с 10% до 64% времени до первого токена: фронтенд каждый раз токенизирует весь транскрипт заново. TokTier делает её stateful, переклеивая только окно вокруг дописанного куска: до 437 раз быстрее токенизатора Hugging Face и минус 16 до 34% медианного TTFT с vLLM.

ai llm inference vllm