Pricing
Every model is priced as a fixed percentage of list price. Prepaid credits, no subscription, no minimum monthly spend.
Claude
80% OF LISTClaude models for Claude Code, Cline, and the Anthropic SDK.
| MODEL | INPUT ours / list | OUTPUT ours / list | CACHED INPUT ours / list | CACHE WRITE ours / list | CONTEXT | YOU SAVE |
|---|---|---|---|---|---|---|
Claude Sonnet 5RECOMMENDED claude-sonnet-5 Best balance of speed and quality. Recommended for most agent work. | $1.60 $2.00 | $8.00 $10.00 | $0.16 $0.20 | $2.00 $2.50 | 1M | −20% |
Claude Opus 4.8 claude-opus-4-8 Flagship coding model for long-horizon, multi-file work. | $4.00 $5.00 | $20.00 $25.00 | $0.40 $0.50 | $5.00 $6.25 | 1M | −20% |
Claude Fable 5 claude-fable-5 Highest capability for the hardest reasoning tasks. | $8.00 $10.00 | $40.00 $50.00 | $0.80 $1.00 | $10.00 $12.50 | 1M | −20% |
Claude Sonnet 4.6 claude-sonnet-4-6 Previous-generation Sonnet. Stable fallback. | $2.40 $3.00 | $12.00 $15.00 | $0.24 $0.30 | $3.00 $3.75 | 1M | −20% |
Claude Haiku 4.5 claude-haiku-4-5 Fastest and cheapest. Good for simple, high-volume calls. | $0.80 $1.00 | $4.00 $5.00 | $0.080 $0.10 | $1.00 $1.25 | 200K | −20% |
Codex
COMING SOONGPT models for Codex, Cursor, and OpenAI-compatible clients.
Not available yet — we are still validating this route. Prices are indicative.
| MODEL | INPUT ours / list | OUTPUT ours / list | CACHED INPUT ours / list | CACHE WRITE ours / list | CONTEXT | YOU SAVE |
|---|---|---|---|---|---|---|
GPT-5.6 TerraRECOMMENDED gpt-5.6-terra Balanced coding model. The default choice for most agent work. | $1.60 $2.00 | $9.60 $12.00 | $0.16 $0.20 | $2.00 $2.50 | 400K | −20% |
GPT-5.6 Sol gpt-5.6-sol Higher capability tier for harder problems. | $3.20 $4.00 | $16.00 $20.00 | $0.32 $0.40 | $4.00 $5.00 | 400K | −20% |
GPT-6 Astra gpt-6-astra Frontier model for the hardest long-horizon work. | $8.00 $10.00 | $40.00 $50.00 | $0.80 $1.00 | $10.00 $12.50 | 400K | −20% |
GPT-5.5LEGACY gpt-5.5 Previous generation. Stable fallback. | $4.00 $5.00 | $24.00 $30.00 | $0.40 $0.50 | $5.00 $6.25 | 400K | −20% |
Prices shown are for requests under 272K input tokens. Above that, the upstream list price steps up and so does ours — the discount stays the same.
GLM
COMING SOONZhipu GLM models — long context, low cost, OpenAI-compatible.
Not available yet — we are still validating this route. Prices are indicative.
| MODEL | INPUT ours / list | OUTPUT ours / list | CACHED INPUT ours / list | CACHE WRITE ours / list | CONTEXT | YOU SAVE |
|---|---|---|---|---|---|---|
GLM-5.3RECOMMENDED glm-5.3 Zhipu flagship. Long context, strong on code, very low cost. | $1.12 $1.40 | $3.52 $4.40 | $0.21 $0.26 | $1.12 $1.40 | 1M | −20% |
GLM-5.3 Flash glm-5.3-flash Fast and cheap. For high-volume, latency-sensitive calls. | $0.12 $0.15 | $0.40 $0.50 | $0.024 $0.030 | $0.12 $0.15 | 1M | −20% |
GLM-5.2 glm-5.2 Previous flagship. Same pricing as 5.3, with a free upstream tier. | $1.12 $1.40 | $3.52 $4.40 | $0.21 $0.26 | $1.12 $1.40 | 1M | −20% |
GLM-5.1LEGACY glm-5.1 Older generation, still capable. Cheapest paid tier upstream. | $1.12 $1.40 | $3.52 $4.40 | $0.21 $0.26 | $1.12 $1.40 | 1M | −20% |
GLM-5LEGACY glm-5 First of the GLM-5 line. Lower list price than 5.1 and up. | $0.80 $1.00 | $2.56 $3.20 | $0.16 $0.20 | $0.80 $1.00 | 1M | −20% |
USD per million tokens. The upper figure is our price; the struck-through figure below it is list price. Cached input is billed at one tenth of input — for coding agents this is usually most of the bill.
Credits
Top up any amount from $20. Credits never expire and work across every model group. What you pay is what you get — no bonus credits or tiers to keep track of.
Working at higher volume? Get in touch about volume pricing.
Before you sign up
GlobalRouter is an independent third-party API gateway. We are not affiliated with, endorsed by, or sponsored by Anthropic. Requests are routed through upstream providers, and model behavior — including system-level instructions and how the model describes itself — can differ from a first-party API. The service is built and tuned for coding agent workloads; evaluate it for your use case before relying on it.
Capacity depends on upstream providers and can be interrupted without notice. We do not offer an uptime SLA. Unused credits are refundable in full within 30 days — see our refund policy.