Per-token input and output rates, context windows, rate limits, and free-tier caps for every AI provider WaxFrame supports. Pick a Builder by cost and capability instead of guessing.
Last updated 2026-06-03. Provider pricing changes a few times a year — use the Check rates link in each provider's row below to verify the current rate before committing to a heavy session.
Per-WaxFrame-round cost assumes ~5,000 input tokens (project setup + reference material + working document + reviewer notes) and ~2,000 output tokens (the Builder's updated draft). That's an average — a 500-word résumé runs lighter; a 4,000-word RFP response with a 10,000-character reference scaffold runs much heavier. Use these numbers for ballpark sizing, not exact billing.
Worker Bee cost is roughly 10–20% of Builder cost because Worker Bees only return numbered suggestions, not full document rewrites. A 6-bee hive's reviewers cost less than the single Builder pass on the same round.
Free-tier caps: "Free tier" means rounds that cost nothing as long as you stay under the provider's daily request quota. Most free tiers also require billing disabled on your account — adding a credit card can silently flip the account to paid-tier routing even on the same model name.
For detailed setup steps on every provider, see the API Key Guide. For a deeper explanation of how tokens are counted and why the Builder needs more headroom than reviewers, see What Are Tokens?.
Default models shipped in WaxFrame. Switch any AI to a newer or cheaper model in-app via Recommend Models.
💡 Billing rates change. Verify each provider's billing page monthly to stay current. Per-round cost is an estimate for a typical WaxFrame Builder pass (~7K total tokens). Click any column header to sort. Default sort: lowest cost per round first.
| Provider | Default model | Input $/M | Output $/M | Context | Max output | Est. $/round | Billing |
|---|
* Gemini free tier only while billing is disabled on your Google AI Studio account. Adding a credit card flips routing to paid tier even on the same model name. See What Are Tokens? for the full caveat.
Token budget is only half the cost equation — providers also rate-limit requests per second (RPS), per minute (RPM), and tokens per minute (TPM). A WaxFrame hive can fire 6+ concurrent calls per round, so per-provider RPS matters.
| Provider | Tier 1 RPM | Tier 1 TPM | Free tier | Notes |
|---|
Tier 1 RPM/TPM figures are typical entry-tier limits for paid accounts. Most providers raise limits automatically as paid usage history grows. Rate-limit headers ("X-RateLimit-Remaining" or equivalent) let WaxFrame surface 429 warnings before they bite — a future release will pipe these into the troubleshooting layer.
The Builder is the highest-token role in WaxFrame — it reads the entire setup, the full document, and every reviewer's notes every round, then writes the updated document back out. Builder choice has the biggest cost impact on a session by a wide margin. Three sensible defaults:
About 30 cents per full hive review. $5 typically covers ~15 full reviews; $10 covers ~30; $20 covers ~60. Most users top up once a quarter, not once a week.
That number is measured, not estimated. v3.63.146 analyzed two recent full 10-AI hive runs (Midea AC review and Manly Bikes TripAdvisor review, both captured 2026-06-04 via DeepDive JSON) using the actual deployed model lineup: gpt-5.5, claude-opus-4-8, gemini-3.1-pro-preview, grok-4.3, sonar-reasoning-pro, mistral-large-latest, deepseek-v4-pro, Qwen3-235B-A22B, command-a-plus-05-2026, jamba-mini-2-2026-01. Per-run cost landed at $0.2997 (Midea) and $0.2984 (Manly Bikes) — both right around the 30¢ mark.
That number drops sharply if you swap any reasoning models for chat-tier alternatives or include Gemini on its free tier. It rises if you run heavier reasoning models throughout. The per-provider rate table above is the source of truth — these numbers move proportionally with provider rate changes.
Gemini Flash on its free tier costs nothing while your Google AI Studio account has billing disabled. Among paid providers, DeepSeek is roughly 10x cheaper per token than Claude or OpenAI for comparable quality, though it responds 60 to 90 seconds slower per round. For paid-tier reliability with the lowest cost, DeepSeek is the value pick; for absolute lowest cost, Gemini Flash free tier wins.
A 4-to-6-AI hive typically costs about 4 to 6 times a single AI per round, since each Worker Bee fires once and the Builder synthesizes. But the hive catches issues a single model misses — factual drift, weak phrasing, missing requirements — because each model has different blind spots. For documents that matter (RFPs, cover letters, business proposals), the multi-AI review usually pays for itself in revision rounds saved.
Gemini's free tier on AI Studio is the only practical free Builder option, and only while billing is disabled. The Builder is the highest-token role in WaxFrame — it reads the project setup, reference material, the full working document, and every reviewer's notes every round, then writes the document back. That's much more than the free Worker Bees use. Free tiers from Mistral, Cohere, and Together AI are useful for Worker Bee roles but tend to hit rate limits as Builder on longer documents.
No — Jamba is Reviewer-only, regardless of plan or pricing tier. AI21's API caps every Jamba response at 4096 output tokens (across Jamba 1.5, 1.6, and 1.7 — Mini and Large). A typical WaxFrame Builder round needs roughly 5K–8K output tokens to emit the refined document plus the required formatting blocks. Jamba truncates before finishing, the round fails, and your API credits are spent on a doomed call. This is an architectural limit on AI21's side, not a setting you can change. WaxFrame filters Jamba out of Builder recommendations automatically and badges it ⚠️ Reviewer-only in the Change Builder model picker so you don't waste credits trying. Jamba still works fine as a Worker Bee (Reviewer role) where ~2K of output is plenty.
A ChatGPT Plus subscription doesn't include API access — the API is billed separately as pay-as-you-go. WaxFrame's value is automated multi-AI fan-out: it sends prompts directly to each AI on your behalf with no copy-paste, no tab switching, no manual orchestration. That requires API keys. The free WaxFrame edition (weirdave.github.io/WaxFrame-Free(opens in a new tab)) runs the same hive workflow but you copy-paste between WaxFrame and each AI's web interface manually.
Provider pricing typically changes a few times a year — most often during product launches or competitive responses. Last updated date is shown above; click the Check rates link in each provider's row to verify the current rate at that provider's billing page before committing to a heavy session. The cost-per-WaxFrame-round figures will move proportionally with input/output rates.
API Key Guide — step-by-step setup for every provider above, with direct links to each console and a "Start Here" walkthrough for first-time API users.
What Are Tokens? — how tokens are counted, why your Builder needs more headroom than reviewers, and how to size your costs before committing to a session.
User Manual — full walkthrough of WaxFrame's setup screens, hive rounds, Builder configuration, and convergence detection.
WaxFrame is fully open source under the AGPL-3.0 license. Every line of code is public and auditable.
💡 Your API keys go directly from your browser to each AI provider. WaxFrame never sees them. Read the source on GitHub(opens in a new tab)
About
Built with ❤️ by WeirDave and Claude.
Enter your license key to continue.
Don't have a key? Buy WaxFrame Pro(opens in a new tab)
Your WaxFrame Pro license is active.
Hive Profiles