AI models on Okou

Claude Fable 5

Anthropic's premium frontier model for the hardest reasoning and long-horizon agentic runs. Always-on thinking, top SWE-bench Pro scores, at a premium above Opus-tier pricing.

Claude Opus 5

Anthropic's flagship for agentic coding and long-horizon work. State of the art on Frontier-Bench and GDPval-AA, at Opus 4.8's $5/$25 pricing.

Claude Opus 4.8

Anthropic's newest Claude 4 flagship. Leads SWE-bench Pro and OSWorld-Verified at release, with dynamic workflows and a 3× cheaper fast mode.

Claude Opus 4.7

Anthropic's flagship Claude 4 model. Highest reasoning quality in the family on multi-step agent loops, long-context recall, and code edits.

Claude Opus 4.6

Anthropic's previous flagship. Same credit cost as Opus 4.7. Keep it pinned only if a specific agent has been validated against this version.

GPT 5.6 Sol

OpenAI's newest flagship reasoning model, in preview on Okou. Top-tier coding and tool-use, routed through the Codex framework.

GPT 5.6 Terra

The balanced workhorse of the GPT-5.6 family, in preview on Okou. Strong coding and tool-use at half the flagship credit cost.

GPT 5.6 Luna

The fast, economy tier of the GPT-5.6 family, in preview on Okou. Built for bulk work and latency-sensitive steps at the lowest GPT-5.6 credit cost.

GPT-5.5

OpenAI's flagship reasoning model. Top-tier coding and tool-use, routed through the Codex framework on Okou.

Claude Sonnet 5

The best balance of speed and intelligence in the Sonnet tier. Near-Opus coding and agentic quality at Sonnet cost, with adaptive thinking on by default.

Claude Sonnet 4.6

The default for most Okou agents. Strong tool-routing accuracy, good long-context behaviour, and the credit baseline. Every other model is priced relative to Sonnet 4.6.

GPT-5.4

OpenAI's workhorse model. ×1 credits, multimodal vision, strong code and tool use — the default GPT-5 for most agents.

GPT-5.4 Mini

OpenAI's cost-saving GPT-5. ×0.3 credits, fast and multimodal — the right pick for bulk pre-filter work on the Codex framework.

DeepSeek V4 Flash

Open-weight 284B MoE built for agent work. 82.7 on Terminal-Bench 2.1, a 1M-token context window, and Okou's cheapest price tier.

GPT Image 2

OpenAI's newer image model for generation and editing, with flexible sizing and stronger prompt adherence than GPT Image 1.

GPT Image 1

OpenAI's text-to-image model. Strong on stylised illustration, character work and editing.

Black Forest Labs' top-tier image model. Highest aesthetic ceiling in the curated lineup — use it for hero shots and editorial-grade output.

ByteDance's SeedDream 4. Photoreal aesthetic at a low per-image cost — the natural default when the brief is a photo.

Veo 3.1 Fast

Google's text-to-video model with native audio. Fast generation, cinematic style and synchronised soundtrack in a single pass.

Kuaishou's Kling V3 4K. Stylised aesthetic, long clips up to 15 seconds, 4K output — the ceiling for editorial and anime-style video.

ByteDance's 30-second video model with native audio and up to 50 multimodal references.

MiniMax H3

MiniMax's multimodal video model with native stereo audio, 2K output and reference-guided generation.

ByteDance's Dreamina Seedance 2.0. Cheapest per-second video in the curated lineup — the right pick for bulk and cost-driven generation.