As of September 26, 2026, GPT-6 Luna and DeepSeek Flash cost least per token. Among $2-input models, Grok 4.7 is cheapest; past 272K tokens, Sonnet 5 beats Sol.
As of November 12, 2026, a 10-second 720p clip costs $0.30 flat on Gemini Omni 1.2 Flash or Muse via LaoZhang API, and up to $1.20 on other video APIs.
As of September 26, 2026, GPT-6 Sol matches Sonnet 5 and Astra matches Fable 5.1 on list price. Prompts over 272K favor Claude; Luna costs a tenth of Haiku 4.5.
Claude Code's login isn't rejected: another process holds the refresh lock in ~/.claude. Retry in a minute; if nothing holds it, remove the empty lock folder.
Claude Code fast mode runs the same Opus 5.5 or Fable 5.2 up to 2.5x faster at 2x the token price, paid from usage credits on plans. Enable it at session start.
There's no separate Fable 5.2 key: any Claude Console key calls claude-fable-5-2 once its organization has API credit. A Pro or Max plan alone doesn't cover it.
Claude Haiku 5.5 costs 1/40 of Opus 5.5 per token on prompts up to 100K and 23–29x less per task at equal effort. Move checkable work; keep agents on Opus.
Claude Code not working in VS Code with command 'claude-vscode.editor.openLast' not found? The extension failed to start: update it and fully restart VS Code.
Claude Opus 5.5 costs $4/$20 per million tokens; Opus 5 through 4.6 stay at $5/$25. Limit resets are free, can't be bought and don't add API credits.
Claude Opus 5.5 lists 60% below GPT-6 Astra per token. On small coding tasks it cost 0.5–0.7x as much; on long agentic work at equal score, about the same.
Codex projects missing after an update or account switch are usually still on disk. Update to 26.1106.51428, fully quit, and back up .codex before re-adding.
Vercel AI Gateway and OpenRouter carry Google's 3–15 s, up-to-4K model. A flat per-clip gateway costs less than per-second billing for 5–10 s clips to 1080p.
Gemini Omni 1.2 Flash ties Seedance 2.5 on text-to-video and edges it on editing at $0.70 vs $2.32 per 10 s clip; Seedance leads image-to-video and 30 s takes.
Run GPT-6 Astra as Claude Code's main model via a local LiteLLM proxy on API billing, or keep Claude in charge and hand tasks to Astra with the Codex plugin.
Set model to gpt-image-2.6-flare or gpt-image-2.6-sunburst; bare gpt-image-2.6 doesn't exist. Both cost the same, and edits without model run on 2.5.
Update to ComfyUI v0.45.0 or later, then link the GIF node's frames and fps outputs to Save Animated WEBP. At most square sizes, 12 frames cost less than 9.
Write GPT Image 2.6 prompts in plain words: result, details, quoted text, exclusions. Quoted text held 2 of 2 times; edits need a size that matches the photo.
Send output_format svg to gpt-image-2.6-flare with an svg.model writer on OpenAI, or to LaoZhang's gpt-image-2.6-web at $0.03 a call with an ordinary key.
As of September 2026, creator plans run about $5–$35 a month and APIs charge $0.04–$0.25 per second at 720p. Sora is gone, and retries set the real cost.
As of November 12, 2026, MiniMax H3 lists at $0.08/s (768P) and $0.13/s (2K); LaoZhang API charges $0.05 and $0.08, so a 5-second 768P clip is $0.25.
At max effort on Artificial Analysis, Claude Opus 5.5 scores 58 to Muse Wonder's 43, and Wonder costs 31% less per task. Wonder Full has no API.
Seedance 2.5 makes 30-second takes for about $0.231 per second at 720p. Veo 3.1, now sold on Google Cloud, adds native 4k and sound from $0.05 per second.
Seedream 5.5 gives real alpha only with Background set to Transparent and PNG output. Solid objects and glow come out clean; glass stayed opaque in tests.
Seedream 5.5 for repeated edits and clean cut-outs; GPT Image 2.6 for GIF, SVG, 4K and see-through glass. Tested November 12, 2026: no 2.6 pixel hit full alpha.