The best free AI model APIs
Dozens of providers now serve real models at $0. Most lists stop there. This one adds the part that matters: independent proof that the model behind each free tier is actually good, ranked on a single benchmark that scores every one of them. The weak free models are left off.
What is the best free AI model API right now?
The best free AI model APIs in 2026 pair a genuinely capable model with a real zero-cost tier. Ranked on the one benchmark that scores them all, the Artificial Analysis Intelligence Index (v4.2): Meta Muse Spark 1.3 leads at 53 of 100, free through the OpenCode Zen Contributor tier, followed by Muse Spark 1.2 and Z AI GLM-5.2. 12 of the 17 ranked models are open-weight, so they survive a free tier being withdrawn. Every model here scores at least 20, so the weak free models are filtered out.
Every free model on one yardstick
The Artificial Analysis Intelligence Index is the one independent benchmark that scores all of these models, so it is the only fair way to rank them together. It is a demanding scale: on v4.2 the best model at any price scores 57, so a free model in the 20s to 50s is genuinely capable.
| Item | Value |
|---|---|
| Muse Spark 1.3 | 53 |
| Muse Spark 1.2 | 47 |
| GLM-5.2 | 43 |
| DeepSeek V4 Flash 0731 | 41 |
| Gemini 3.5 Flash | 38 |
| MiniMax M3 | 36 |
| Hy3 | 33 |
| Inkling | 32 |
| Inkling Small | 32 |
| Nemotron 3 Ultra | 30 |
| MiMo V2.5 | 30 |
| MiMo V2 Omni | 28 |
| Ling 3.0 Flash | 27 |
| LongCat 2.0 | 26 |
| MiMo V2 Flash | 26 |
| Ring 2.6 1T | 24 |
| Gemma 4 31B | 22 |
The free models, ranked
Sort by intelligence or context window, and filter to open-weight models. Coding is 20% of the index itself, carried by Terminal-Bench v2.1 and SciCode, so it is already folded into the ranking. Each row links to the providers that serve it free.
Good free models, ranked by intelligence
Every model here clears the quality bar (Artificial Analysis Intelligence Index of at least 20) and is callable at $0 on at least one hosted provider. Sort, or show only open-weight models.
| # | Model | AA Index | Context | Free via |
|---|---|---|---|---|
| 1 | Muse Spark 1.3 Meta | 53 | 1M | OpenCode Zen |
| 2 | Muse Spark 1.2 Meta | 47 | 1M | OpenCode Zen |
| 3 | GLM-5.2 Z AI· open weight | 43 | 256K | OpenRouter |
| 4 | DeepSeek V4 Flash 0731 DeepSeek· open weight | 41 | 200K | OpenCode Zen |
| 5 | Gemini 3.5 Flash Google | 38 | 1M | Google AI Studio |
| 6 | MiniMax M3 MiniMax· open weight | 36 | 1M | OpenCode ZenOpenRouter |
| 7 | Hy3 Tencent· open weight | 33 | 190K | OpenCode Zen |
| 8 | Inkling Thinking Machines· open weight | 32 | 1M | OpenRouter |
| 9 | Inkling Small Thinking Machines· open weight | 32 | 1M | OpenRouter |
| 10 | Nemotron 3 Ultra NVIDIA· open weight | 30 | 1M | OpenCode ZenOpenRouterNVIDIA NIM (build.nvidia.com) |
| 11 | MiMo V2.5 Xiaomi· open weight | 30 | 200K | OpenCode Zen |
| 12 | MiMo V2 Omni Xiaomi· open weight | 28 | 262K | OpenCode Zen |
| 13 | Ling 3.0 Flash InclusionAI | 27 | 262K | OpenCode Zen |
| 14 | LongCat 2.0 LongCat | 26 | 1M | OpenCode Zen |
| 15 | MiMo V2 Flash Xiaomi· open weight | 26 | 262K | OpenCode Zen |
| 16 | Ring 2.6 1T InclusionAI· open weight | 24 | 262K | OpenCode Zen |
| 17 | Gemma 4 31B Google· open weight | 22 | 262K | OpenRouter |
AA Index is the Artificial Analysis Intelligence Index (0-100, v4.2): one independent score that blends reasoning, knowledge and coding across ten evaluations, so it ranks every model on the same yardstick. The scale runs low: the best model at any price scores 57. Coding is 20% of it, carried by Terminal-Bench v2.1 and SciCode. Where a lab ships several reasoning-effort settings, the figure is the highest one Artificial Analysis publishes, named in each model's note. Free-via links open each provider; verify current limits at the source, as free tiers change often.
The free tiers, and their limits
The same model is often free on several providers, each with its own catch. Here is what each free tier actually gives you, verified against the provider source on September 5, 2026. Free tiers change often: confirm at the link before you build on one.
| Provider | Free tier | Catch |
|---|---|---|
| Google AI Studio | Gemini Flash and Flash-Lite free (Pro tiers left the free tier on 2026-04-01); roughly 5-15 req/min and 20-1,500 req/day depending on model | Inputs may be used to improve the model. |
| OpenRouter | 22 models tagged :free at 20 req/min and 50 req/day (1,000/day after a one-time $10 credit purchase) | Account signup only. |
| OpenCode Zen | 31 models free through the OpenCode CLI and Desktop, including Muse Spark 1.3 and 1.2, MiMo V2.5, Nemotron 3 Ultra, MiniMax M3, DeepSeek V4 Flash and the Big Pickle stealth model | Inputs may be used to improve the model. |
| NVIDIA NIM (build.nvidia.com) | Open models free at about 40 req/min after phone verification | Phone verification. |
| GroqFastest-class inference on custom LPU hardware | Open models free at roughly 1,000 req/day on larger models, up to 14,400 on small ones; 12K tokens/min | Account signup only. |
| CerebrasAmong the fastest output speeds available | Open models free at 30 req/min, 14,400 req/day, 60K tokens/min | Account signup only. |
| Cloudflare Workers AI | 10,000 neurons/day free across the model catalog (a usage credit, not a request cap) | Account signup only. |
How this list is built
Two grounded inputs. First, which providers serve a model at $0, checked against each provider's own rate-limit documentation and the live OpenRouter API, and dated. Second, how good that model actually is, measured by the Artificial Analysis Intelligence Index, read from the public Artificial Analysis leaderboard. This site does not run the evaluations.
One benchmark, so the ranking is fair
Earlier versions of this page mixed coding benchmarks because no single coding test scored every model. The fix is to rank on the Artificial Analysis Intelligence Index, the one independent benchmark that scores all of them. It is a single 0-100 number that blends ten evaluations across four weighted categories: Agents 30%, Coding 20% (Terminal-Bench v2.1 and SciCode), Scientific Reasoning 20% and General 30%. Coding ability is already folded into the figure and every model is judged on the same yardstick. Artificial Analysis stopped publishing the individual evaluation scores on its public leaderboard with v4.2, so this page no longer breaks Terminal-Bench and SciCode out as their own columns rather than showing figures it cannot cite. The quality gate is the index is demanding: on v4.2 the best model of any price scores 57 and the strongest free model here scores 53, so the 0-100 scale runs low. A free model scoring at least 20 is genuinely capable. The v4.2 recalibration pushed several previously listed free models below the bar, and they are no longer ranked: OpenAI gpt-oss-120b (16), NVIDIA Nemotron 3 Super (19), Google Gemma 4 26B A4B (19), NVIDIA Nemotron 3.5 Lightning (16) and Alibaba Qwen3 Next 80B A3B (8). They are still free; they are just no longer models we would tell you to build on.
Why "free" has a catch
A free API tier is a customer-acquisition cost for the provider, not charity. That shapes the three catches in the table above: rate limits (requests per minute and per day, often tightened over time),data use (several free tiers may train on your inputs, so never send confidential data), and availability (community-funded free pools can be throttled without notice). For anything beyond prototyping or low-volume personal use, price the paid tier with the cost-per-task calculatorbefore you depend on a free one. The OpenRouter AI guideexplains that pool's daily limits, paid-credit fee and routing tradeoffs.
Most of these free models are open-weight, part of a wider shift: open-source LLMs are overtaking closed models, and many are now good enough to run locally if a hosted free tier ever disappears.
Free, but not ranked
A model is ranked only once the Artificial Analysis index scores it. These are offered free but not yet scored on that benchmark, so they are not ranked here:
- Big Pickle (OpenCode Zen). A stealth coding model OpenCode is running free while it collects feedback. No lab, no published weights and no benchmark score, so it cannot be ranked.
- Ox Alpha (Unlimited) (OpenCode Zen). A stealth model offered free with no stated request cap. Artificial Analysis does not score it.
- Grok Code Fast 1 (OpenCode Zen). SpaceXAI coding model, free on OpenCode Zen. Artificial Analysis scores the Grok 4.x line but not this coding-specific model.
- Laguna S 2.1 and Laguna XS 2.1 (OpenCode Zen / OpenRouter). poolside open-weight coding models, free on both gateways, but Artificial Analysis does not score either.
- Ling 3.0 Flash Fin and Ling 3.0 Flash Sante (OpenCode Zen / OpenRouter). Finance-tuned and health-tuned Ling 3.0 Flash variants. Artificial Analysis scores only the base Ling 3.0 Flash, so the tuned variants cannot be ranked separately.
- MiniMax M2.1, M2.5 and M2.7 (OpenCode Zen / OpenRouter). Earlier MiniMax releases still offered free. Artificial Analysis v4.2 scores M3 but not these.
- GLM-4.7 and GLM-5 (OpenCode Zen). Earlier Z AI models free on OpenCode Zen. v4.2 scores GLM-5.2 and GLM-5.3, but not these two.
- Kimi K2.5 (OpenCode Zen). A Moonshot open-weight model free on OpenCode Zen. v4.2 scores K3 and K2.7 Code, not K2.5.
- North Mini Code (OpenCode Zen / OpenRouter). A Cohere coding model free on both gateways. Artificial Analysis scores it at 13 on v4.2, below this page quality gate, so it is listed here rather than ranked.
For per-token rates once you outgrow the free tiers, see the AI model release tracker, or rank paid models by value on the value leaderboard. For when each of these shipped and what it charged on day one, see AI model releases by month.
Know a free tier that is missing, or one whose limits have changed? Submit it or send a correction. Listings are free, editorial, and cannot be bought.
Frequently asked questions
- What is the best free AI model API in 2026?
- For raw quality, Muse Spark 1.3 leads at 53 of 100 on the Artificial Analysis Intelligence Index, free through the OpenCode Zen Contributor tier. Read the terms before you use it: that tier grants permission to train future Meta models on your prompts and completions. For an open-weight model you can also self-host, GLM-5.2 is the strongest at 43. For a first-party free tier from a frontier lab, Gemini 3.5 Flash on Google AI Studio is the only one here. The best one depends on whether you optimize for capability, open weights, or who gets to keep your data.
- Are free AI models good enough for real work?
- Increasingly, yes. The top free model here, Muse Spark 1.3, scores 53 of 100 on the Artificial Analysis Intelligence Index, a demanding scale where the best model at any price scores 57. So the leading free models sit within a few points of the frontier. The real limits are throughput and terms rather than quality: free tiers cap requests per minute and per day, and several of them take your prompts as payment, so they suit prototyping, low-volume tools and personal use more than high-traffic production or anything confidential.
- What is the catch with free AI model APIs?
- Three things. First, rate limits: most free tiers cap requests per minute and per day (often tightened over time, as Groq did in 2026). Second, data use: several free tiers (Google AI Studio outside the EEA, the OpenCode Zen free models) may use your inputs to improve their models, so do not send confidential data. Third, availability: community-funded free pools such as OpenRouter free models can be throttled or rotated without notice.
- Which of these free models are open-weight?
- 12 of the 17 ranked models are open-weight, meaning you can also download and self-host them: Z AI GLM-5.2, DeepSeek V4 Flash, MiniMax M3, Tencent Hy3, both Thinking Machines Inkling models, NVIDIA Nemotron 3 Ultra, three Xiaomi MiMo models, InclusionAI Ring 2.6 1T and Google Gemma 4 31B. The closed exceptions are the two Meta Muse Spark models and Google Gemini 3.5 Flash, plus InclusionAI Ling 3.0 Flash and LongCat 2.0. Open weights matter most for free use, because if a hosted free tier disappears the model itself does not.
- How are these free models graded?
- On a single benchmark that scores every model here: the Artificial Analysis Intelligence Index (v4.2). It is one 0-100 number blending ten independent evaluations across four weighted categories, Agents 30%, Coding 20%, Scientific Reasoning 20% and General 30%, so one figure ranks the whole table on the same yardstick. Scores are read from the public Artificial Analysis leaderboard; this site does not run the evaluations. A model is listed only if it scores at least 20. Note that v4.2 replaced v4.1 in September 2026 and is not comparable with it: it added two evaluations, dropped GPQA Diamond and rebalanced every weight, which moved most scores down by several points. The figures on this page were re-read against v4.2 on 2026-09-05.
Sources
- Artificial Analysis (2026). Artificial Analysis Intelligence Index (v4.2). Scores read 2026-09-05 from the public leaderboard. https://artificialanalysis.ai/leaderboards/models
- Google AI Studio (2026). Free-tier terms and rate limits. Verified 2026-06-28. https://ai.google.dev/gemini-api/docs/rate-limits
- OpenRouter (2026). Free-tier terms and rate limits. Verified 2026-09-05. https://openrouter.ai/docs/api-reference/limits
- OpenCode Zen (2026). Free-tier terms and rate limits. Verified 2026-09-05. https://models.dev/api.json
- NVIDIA NIM (build.nvidia.com) (2026). Free-tier terms and rate limits. Verified 2026-06-28. https://build.nvidia.com/
- Groq (2026). Free-tier terms and rate limits. Verified 2026-06-28. https://console.groq.com/docs/rate-limits
- Cerebras (2026). Free-tier terms and rate limits. Verified 2026-06-28. https://inference-docs.cerebras.ai/support/pricing
- Cloudflare Workers AI (2026). Free-tier terms and rate limits. Verified 2026-06-28. https://developers.cloudflare.com/workers-ai/platform/pricing/
- OpenRouter (2026). Models API (used to verify which models are currently $0). https://openrouter.ai/api/v1/models
Machine-readable data: /free-ai-models.json. Methodology and the benchmark source are documented in the site repo.