BlockRun
Get started

Available Models

Pay with USDC on Base. No account needed.

New · July 2026Claude Opus 5GPT-5.6 Sol · Terra · LunaClaude Sonnet 5OpenAI Sora 2

GPT-5.6 is live — three tiers: Sol, Terra, and Luna.

Chat: OpenAI openai/gpt-5.6-sol — flagship tier, deepest reasoning for complex coding and long-horizon agentic work, at $5 in / $30 out per 1M. openai/gpt-5.6-terra covers everyday coding and agentic tasks at $2 in / $12 out, and openai/gpt-5.6-luna handles high-volume, latency-sensitive work at $0.20 in / $1.20 out — all with 1M context, 128K output, and vision. Anthropic anthropic/claude-sonnet-5 — near-Opus coding and agentic quality at Sonnet pricing ($3 in / $15 out) — and the new flagship anthropic/claude-opus-5, a step-change over Opus 4.8 at the same price ($5 in / $25 out), round out the lineup. Video: OpenAI sora-2 (via Azure) at $0.10/sec — 720p with synced audio, 4/8/12s clips — alongside ByteDance Seedance (seedance-1.5-pro ≈ $0.493/5s · seedance-2.0-fast ≈ $1.276/5s · seedance-2.0 ≈ $1.595/5s) and xAI Grok Imagine at $0.05/sec. For consistent characters across clips, enroll a Seedance Virtual Portrait ($0.011, AI character) or Seedance RealFace ($0.011, real person + 1-min on-phone liveness check, no KYC).

Generated assets are mirrored to BlockRun's storage so URLs don't expire. Responses return a permanent url plus the upstream source_url. See the Image Generation and Video Generation docs.

92 models

How pricing works on BlockRun

Every model above is billed per request, in USDC, settled on-chain through the x402 protocol. There is no subscription, no monthly minimum, and no API key to provision — a request arrives unpaid, the gateway answers with a signed price quote, your client signs it, and the retry returns the completion. Chat models are priced per million input and output tokens; image, video, music and speech models are priced per image, per second, per track, and per thousand characters respectively.

Token prices track each upstream lab's public list price, plus a flat platform margin and a small per-transaction fee that covers on-chain settlement. Because you pay per call, an agent that makes ten requests a day costs cents a month, and one that makes ten thousand pays exactly ten thousand times the per-call price — the unit economics do not change with volume, and nothing expires unused.

Choosing a model

The catalog spans 71 chat and reasoning models alongside image, video, music and speech generation, for 92 in total. Use the Reasoning filter for long-horizon planning and hard analytical work, Coding for repository-scale edits and agentic tool use, and Vision when requests carry images. The Free filter lists models that need no payment header at all — useful for prototyping an integration before you fund a wallet.

Every model is reachable through the same OpenAI-compatible /v1/chat/completions endpoint, so switching between them is a one-line change to the model field. See the documentation for request shapes and the pricing page for a full per-model rate card.