Intelligence score vs. live price across 6 models: how much intelligence you get per dollar.
| # | Model | Intelligence | Blended $ | Score / $ | Context |
|---|---|---|---|---|---|
| 1 | GLM 5.3 Flashz-aiopen weights | 41.8 | $0.0440.029/0.092 | 939.3 | 1.0M |
| 2 | Ling 3.1 Flashinclusionai | 41.1 | $0.1130.075/0.225 | 365.3 | 1.0M |
| 3 | MiMo-V2.6-Proxiaomiopen weights | 46.3 | $0.5440.435/0.870 | 85.2 | 1.1M |
| 4 | GLM 5.3z-aiopen weights | 44.8 | $0.9600.080/3.60 | 46.7 | 1.0M |
| 5 | Step 5 Previewstepfun | 43.7 | $1.431.00/2.70 | 30.7 | 1.0M |
| 6 | Muse Spark 1.3meta | 45.1 | $2.001.25/4.25 | 22.6 | 1.0M |
Zero-priced on OpenRouter — free daily quota applies (20 req/min; 50–1000 req/day).
Extracted from official pages by a deterministic script — no LLM involved. Changes are detected and flagged daily; entries marked “manual” are human-verified with a date.
60 requests/min and 1,000 requests/day with personal Google account
A personal Google account is enough, including the Gemini 3 series and a 1M context window
Free plans including Copilot Free and Copilot Student · Limited to 2000 completions per month on Copilot Free · Pro $10 USD per month · Pro+ $39 USD per month · Max $100 USD per month
The free tier is directly usable (includes completion and chat quota); students / teachers / popular open-source maintainers get Pro features for free
Limited agent requests
The Hobby free tier requires no credit card; there is also a Students plan (cursor.com/students)
Free enterprise activation, free personal use
Free for personal use; free activation for enterprise (includes coding agent)
Free $0: Light quota to code with agents, limited model selection, unlimited inline edits and Tab completion
Official pricing page is JS-rendered; needs headless browser verification (script only tracks verification date)
No free tier: Claude chat has Free $0, but Claude Code requires Pro / Max / API
Official pricing page only has a $0 free tier for the chat product; fake "Claude Code forever free" entries have appeared
Official docs do not list a fixed free quota (pay-as-you-go / account login required); numbers to be added after manual verification
Avoid inventing "completely free" like older entries did
International personal users can manually claim 100 Credits per day on desktop (one claim per round, valid for 30 days); some models (e.g. Qwen3.8-Flash) were once free to call with a limited-time 0 billing coefficient
Daily claim limit, unused credits do not accumulate, requires manual action on desktop; activity validity subject to official docs
Offers education discount: teachers / students get free or discounted quota after verification (see post-verification page)
Requires education identity verification; discount tiers and regions may change, subject to the official education section
API and web offer low-cost tiers long-term; events like "0 credits" and "V3 free calls" return periodically—watch official announcements
The 0 credits event is a limited-time return, not permanent; refer to the official pricing page / announcements in real time
Most mainstream models offer a free tier (model pages marked Free / $0 pricing), commonly used for trial runs and evaluation, with no or very low billing
Free-tier models and quota are subject to official adjustments; for evaluation / trial runs, prioritize models marked Free
Google AI Studio offers a free tier: free quota for Gemini series models, requires a Google account
Free tier has rate limits; domestic access requires a compliant network
La Plateforme offers free-tier model calls (e.g., mistral-small), free quota available upon registration
Free quota and models are subject to official adjustments
Use Grok for free via X (Twitter) (limited quota); API is pay-as-you-go
Free tier has a daily query limit
Offers a free tier: limited daily Q&A queries (Pro tier unlocks advanced models and higher quota)
Free-tier models and usage counts are limited
OpenAI offers a ChatGPT free tier (limited quota and models), usable upon registration
Free-tier features / models are limited; API is billed separately on a pay-as-you-go basis
⚠ 1 source(s) failed this run and kept the previous value: Gemini CLI
Same model, different inference providers on OpenRouter. Cheaper endpoints are often quantized (fp8/fp4) — check the quant tag.
Method: intelligence score (Artificial Analysis Intelligence Index) divided by blended price (input 3 : output 1, USD per 1M tokens, OpenRouter list price). Data is collected deterministically every day; no model inference is involved.