Resources
The Token Governor
Price data, model cost guides, loop patterns and margin playbooks for teams who pay for LLM tokens and sell software on top of them.
- Per call
- $0.0466
- Per 100k calls
- $4,662
- Cache saves
- $3,888
(9,600 x $3.00 + 14,400 x $0.30 + 900 x $15.00) / 1M
Compare all models in the index- models priced in the index
- 36models priced in the index
- per-model cost guides
- 28per-model cost guides
- FinOps guides
- 15FinOps guides
- engineering and FinOps posts
- 12engineering and FinOps posts
Flagship dataset
The LLM Price Index
Input, output, cached input and cache write prices per 1M tokens for 36 models, side by side. Licensed CC BY 4.0, last full review June 2026.
- Sort and filter by provider and tier
- Cache savings calculator per model
- Blended cost at a 3:1 input to output ratio
| Model | Input | Output | Cached | Blended |
|---|---|---|---|---|
 Mistral Large 2 | $2.00 | $6.00 | n/a | $3.00 |
 GPT-5 | $1.25 | $10.00 | $0.13 | $3.44 |
 Gemini 2.5 Pro | $1.25 | $10.00 | $0.31 | $3.44 |
| GPT-4.1 | $2.00 | $8.00 | $0.50 | $3.50 |
| GPT-4o | $2.50 | $10.00 | $1.25 | $4.38 |
 Command R+ | $2.50 | $10.00 | n/a | $4.38 |
Model cost guides
Every major model, priced and explained
Price breakdowns, what drives cost for each model, circuit breaker settings and a starting PALM SOUQ policy.
All model guides
DeepSeek
Amazon
Agent loop taxonomy
A catalog of the recursion traps that turn agents into invoices: tool thrashing, retry storms, planner and executor ping-pong, sub-agent fan-out, re-summarize loops. Each with its signature, the heuristic that catches it and the cost math.
HTTP/1.1 429 Too Many Requests X-PalmSouq-Reason: circuit_breaker_triggered X-PalmSouq-Heuristic: tool_thrashingBrowse the taxonomy
Runaway cost calculator
Model, steps per session and tenants in; monthly loop exposure out, with the formulas shown.
Glossary
Cache write, token velocity, fail-open, blended cost, burn rate and the rest of the LLM FinOps vocabulary, defined.
FinOps guides
All 15- Normalizing token usage across OpenAI, Anthropic, Gemini and BedrockAttribution
- Credit wallets and prepaid drawdown for AI productsPricing
- Cost anomaly detection for LLM spend: building per-tenant baselinesReporting
- Unit economics dashboards for AI SaaS finance teamsReporting
- Forecasting LLM spend from usage drivers, not last month's invoiceForecasting
- Fail-open vs fail-closed for LLM gateways: a decision frameworkReliability
From the blog
All 12- What 25 power users do to a $49,500 MRR support SaaS14 Oct 2026 · Ananya Iyer · 3 min
- Under 5 ms: the latency budget of an in-path LLM gateway9 Sept 2026 · Rohan Malhotra · 3 min
- Per-tenant rate limits are not per-tenant budgets5 Aug 2026 · Vineet Saresar · 3 min
- Introducing the PALM SOUQ LLM Price Index8 Jul 2026 · Ananya Iyer · 3 min
Newsletter
Get The Token Governor every week
Price changes from the index the week they happen, one loop teardown or margin example with the arithmetic, and new guides. Written by the PALM SOUQ FinOps research and engineering teams.
- Weekly, one email
- Unsubscribe in one click
- No third-party sponsors
- Prices sourced from the index
Don't wait for a surprise $10,000 OpenAI invoice.
Route your agent traffic through PALM SOUQ in 2 minutes. Change one base URL, pass a tenant header, and every token is attributed and governed.
14-day free production trial · 100k requests included · no card required