anytimelime — the free AI field report

2026-09-20 · generated by a free model interviewing free models · anytimelime, formerly Project42 · get the engine →

What's free in AI right now — 2026-09-20

The morning started as a low‑churn affair, then five models flipped on their tool‑use switch for the first time—zai-org/GLM-5.3-Flash, tokeness/free, nex-agi/n2.5-pro:free, nvidia/ising-calibration-1.5-31b and nvidia/nemotron-3-nano-omni-30b-a3b-reasoning—while half a dozen that were handy yesterday went quiet. The most striking surprise is the paywall crawl: six SambaNova models (DeepSeek-V3.2, Meta-Llama-3.3-70B-Instruct, MiniMax-M2.7, MiniMax-M3, gemma-4-31B-it and gpt-oss-120b) are now behind a payment gate, turning what looked like free experiments into a billing prompt.

What feels like a live free‑tier right now

The standout free, tool‑capable models are a tight group of sub‑second workhorses and a few heavier but still usable options. The three inclusionai flashes—ling-3.0-flash-fin:free (0.7 s), ling-3.0-flash-sante:free (0.8 s) and ling-3.0-flash-vl:free (0.9 s)—still deliver fast, structured replies for finance, health and vision tasks. cohere/north-mini-code:free (0.7 s) is the go‑to coder, while dots-studio/dots-3-note-preview:free (1.7 s) handles concise notes. If you need a tool‑ready heavyweight, nvidia/nemotron-3-nano-omni-30b-a3b-reasoning (5.2 s) and the fresh‑off‑the‑press nvidia/ising-calibration-1.5-31b (15.7 s) are the only models that both accept tools and stay free. The OpenRouter wrappers like liquid/lfm-2.5-2.6b:free (1.1 s) and openrouter/free (1.5 s) round out the usable set.

The losses are a quieter echo: qwen/qwen3.8-27b:free, nvidia/nemotron-3-super-120b-a12b:free, poolside/laguna-xs-2.1:free, mistralai/mistral-nemotron and nvidia/nemotron-3.5-lightning-30b-a3b have all dropped off today’s tool list, leaving only the fresh debuts to pick up the slack. The six SambaNova models that once flirted with free status are now blocked behind a payment wall, so any attempt to probe them hits a billing prompt instead of a tool response.

Running it yourself

Clone github.com/fusionpoweredsoftware/anytimelime, set LIME_KEY to an OpenRouter key, and run ./lime.sh. It pulls today’s verified roster and rotates across the passing models, never exposing your key to the outside world.

Today's passing roster

The measured table

Click any model for its endpoint and ready-to-run snippets in curl, Python, Go, Rust, and more.

latencymodelvendortool callfree nowagenticyesterday3 days agoa week ago
0.2hunyuan-largeHY-Chat (HF Space)untestedDOWN—···
4.0zai-org/GLM-5.3-FlashGonka BrokeryesFREE3/5···
4.9tokeness/freeTokenessyesFREE3/5···
0.2gpt-oss-20bhf-openai-proxy (HF Space)untestedDOWN—···
0.2meganova-ai/manta-flash-1.0MegaNova AIuntestedDOWN—·✗✗
1.1openai-fastPollinationsuntestedDOWN—···
0.3aisingapore/Gemma-SEA-LION-v4-27B-ITSEA-LIONuntestedDOWN—·✗✗
0.201-ai/yi-largenvidiauntestedDOWN—✗✗✗
0.7cohere/north-mini-code:freeopenrouteryesFREE3/5✓✓✓
0.4DeepSeek-V3.1sambanovauntestedDOWN—✗✗✗
0.2adept/fuyu-8bnvidiauntestedDOWN—✗✗✗
1.7dots-studio/dots-3-note-preview:freeopenrouteryesFREE3/5✓✓✓
0.3DeepSeek-V3.2sambanovauntestedCARD—✗✗✗
0.2ai21labs/jamba-1.5-large-instructnvidiauntestedDOWN—✗✗✗
0.3google/gemma-4-26b-a4b-it:freeopenrouteryes (past)DOWN—✗✗✗
0.3Meta-Llama-3.3-70B-InstructsambanovauntestedCARD—✗✗✗
0.1aisingapore/sea-lion-7b-instructnvidiauntestedDOWN—✗✗✗
0.3google/gemma-4-31b-it:freeopenrouteryes (past)DOWN—✓✗✗
0.3MiniMax-M2.7sambanovauntestedCARD—✗✗✗
0.1bigcode/starcoder2-15bnvidiauntestedDOWN—✗✗✗
0.7inclusionai/ling-3.0-flash-fin:freeopenrouteryesFREE3/5✓✓✓
0.3MiniMax-M3sambanovauntestedCARD—✗✗✗
0.1databricks/dbrx-instructnvidiauntestedDOWN—✗✗✗
0.8inclusionai/ling-3.0-flash-sante:freeopenrouteryesFREE3/5·✓✓
0.3gemma-4-31B-itsambanovauntestedCARD—✗✗✗
0.2deepseek-ai/deepseek-coder-6.7b-instructnvidiauntestedDOWN—✗✗✗
0.9inclusionai/ling-3.0-flash-vl:freeopenrouteryesFREE3/5·✓✓
0.3gpt-oss-120bsambanovauntestedCARD—✗✗·
60.1deepseek-ai/deepseek-v4-flash-0731nvidiayes (past)DOWN—✓✗·
1.1liquid/lfm-2.5-2.6b:freeopenrouteryesFREE3/5✓✓·
0.2google/codegemma-1.1-7bnvidiauntestedDOWN—·✗·
0.9nex-agi/nex-n2.5-mini:freeopenrouteryesFREE3/5·✓·
0.1google/codegemma-7bnvidiauntestedDOWN—·✗·
20.7nex-agi/nex-n2.5-pro:freeopenrouteryesFREE3/5·✗·
0.2google/deplotnvidiauntestedDOWN—·✗·
0.4nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:freeopenrouteruntestedDOWN—·✗·
0.6google/diffusiongemma-26b-a4b-itnvidiayesFREE3/5·✓·
0.4nvidia/nemotron-3-super-120b-a12b:freeopenrouteryes (past)DOWN—·✓·
0.1google/gemma-2bnvidiauntestedDOWN—·✗·
25.0nvidia/nemotron-3-ultra-550b-a55b:freeopenrouteryesFREE3/5·✓·
0.1google/gemma-3-12b-itnvidiauntestedDOWN—·✗·
0.1nvidia/nemotron-3.5-content-safety:freeopenrouteruntestedDOWN—·✗·
0.1google/gemma-3-4b-itnvidiauntestedDOWN—·✗·
6.5nvidia/nemotron-3.5-lightning:freeopenrouteryesFREE3/5·✓·
35.7google/gemma-4-31b-itnvidiayesFREE3/5·✓·
1.5openrouter/freeopenrouteryesFREE3/5·✓·
0.1google/recurrentgemma-2bnvidiauntestedDOWN—·✗·
16.4poolside/laguna-s-2.1:freeopenrouteruntestedDOWN—·✗·
0.2ibm/granite-3.0-3b-a800m-instructnvidiauntestedDOWN—·✗·
0.2poolside/laguna-xs-2.1:freeopenrouteryes (past)DOWN—·✓·
0.2ibm/granite-3.0-8b-instructnvidiauntestedDOWN—·✗·
0.2qwen/qwen3.8-27b:freeopenrouteryes (past)DOWN—·✓·
0.2ibm/granite-34b-code-instructnvidiauntestedDOWN—·✗·
0.1z-ai/glm-5.2:freeopenrouteruntestedDOWN—·✗·
0.1ibm/granite-8b-code-instructnvidiauntestedDOWN—·✗·
0.1meta/codellama-70bnvidiauntestedDOWN—·✗·
1.0meta/llama-3.2-11b-vision-instructnvidiayesFREE3/5·✓·
60.1meta/llama-3.2-90b-vision-instructnvidiauntestedDOWN—·✗·
0.1meta/llama2-70bnvidiauntestedDOWN—·✗·
2.1meta/muse-glimmer-30bnvidiayesFREE3/5·✓·
0.2microsoft/kosmos-2nvidiauntestedDOWN—·✗·
0.2microsoft/phi-3-vision-128k-instructnvidiauntestedDOWN—·✗·
0.2microsoft/phi-3.5-moe-instructnvidiauntestedDOWN—·✗·
0.2mistralai/codestral-22b-instruct-v0.1nvidiauntestedDOWN—·✗·
0.1mistralai/mistral-7b-instruct-v0.3nvidiauntestedDOWN—·✗·
0.2mistralai/mistral-largenvidiauntestedDOWN—·✗·
0.2mistralai/mistral-large-2-instructnvidiauntestedDOWN—·✗·
60.1mistralai/mistral-nemotronnvidiayes (past)DOWN—·✓·
0.1mistralai/mixtral-8x22b-v0.1nvidiauntestedDOWN—·✗·
0.1moonshotai/kimi-k2.6nvidiauntestedDOWN—·✗·
60.0moonshotai/kimi-k3nvidiauntestedDOWN—·✗·
0.1nv-mistralai/mistral-nemo-12b-instructnvidiauntestedDOWN—·✗·
0.2nvidia/ai-synthetic-video-detectornvidiauntestedDOWN—·✗·
0.1nvidia/cosmos-reason2-8bnvidiauntestedDOWN—·✗·
15.7nvidia/ising-calibration-1.5-31bnvidiayesFREE3/5·✓·
0.1nvidia/llama-3.1-nemotron-51b-instructnvidiauntestedDOWN—·✗·
0.1nvidia/llama-3.1-nemotron-70b-instructnvidiauntestedDOWN—·✗·
0.1nvidia/llama-3.1-nemotron-ultra-253b-v1nvidiauntestedDOWN—·✗·
0.1nvidia/llama3-chatqa-1.5-70bnvidiauntestedDOWN—·✗·
0.1nvidia/mistral-nemo-minitron-8b-8k-instructnvidiauntestedDOWN—·✗·
5.2nvidia/nemotron-3-nano-omni-30b-a3b-reasoningnvidiayesFREE3/5·✗·
1.1nvidia/nemotron-3-super-120b-a12bnvidiayesFREE3/5·✓·
1.7nvidia/nemotron-3-ultra-550b-a55bnvidiayesFREE3/5·✓·
0.1nvidia/nemotron-3.5-content-safetynvidiauntestedDOWN—·✗·
60.0nvidia/nemotron-3.5-lightning-30b-a3bnvidiayes (past)DOWN—·✓·
0.1nvidia/nemotron-4-340b-instructnvidiauntestedDOWN—·✗·
0.1nvidia/nemotron-4-340b-rewardnvidiauntestedDOWN—·✗·
0.2nvidia/nemotron-nano-3-30b-a3bnvidiauntestedDOWN—·✗·
0.2nvidia/nemotron-parsenvidiauntestedDOWN—·✗·
0.2nvidia/nemotron-parse-2.0nvidiauntestedDOWN—·✗·
0.2nvidia/neva-22bnvidiauntestedDOWN—·✗·
0.1nvidia/nvclipnvidiauntestedDOWN—·✗·
0.1nvidia/riva-translate-4b-instructnvidiauntestedDOWN—·✗·
1.0nvidia/riva-translate-4b-instruct-v1.1nvidiauntestedDOWN—·✗·

21 made a tool call, 0 answered in text only, of 94 discovered. A fail only means the endpoint wasn't free to use right then — it demanded a payment method, or it was down or erroring — so tool-call ability stayed untested, not disproven; a model that passed in a past report keeps its yes. Numbers are single-run measurements from one location — your mileage will vary. Prose by cohere/north-mini-code:free; the table is measured, never model-written. Reproduce with ./free-scan.sh from the engine repo.