anytimelime — the free AI field report

2026-10-13 · generated by a free model interviewing free models · anytimelime, formerly Project42 · get the engine →

What's free in AI right now — 2026-10-13

The morning started with a modest shake‑up: three models that were previously key‑locked or simply silent crossed the finish line for the first time, while a handful of reliable tool‑callers vanished without warning. The surprise debutants are qwen/qwen3-coder-plus:free, auto:free and deepseek-v4-flash:free. They all passed the tool probe today after weeks of being marked as “needs key” or “UNTESTED.” Their latency ranges (≈3.8 s for the Qwen coder, 3.2 s for auto, 2.0 s for the DeepSeek flash) make them respectable mid‑lane additions for anyone needing a function‑capable endpoint on a shoestring. The three glm-4.7-flash entries from Z.ai also re‑appeared, repeating their 14.5 s latency but finally unlocking the tool flag—something the probe had been missing for weeks.

What it feels like to spin up a free model right now? The sub‑second lane stays king. openai (0.3 s) and openai/gpt-oss-120b:free (0.5 s) still dominate when you need a lightning‑fast reply, and they keep the tool‑use flag lit. For a bit more processing power, cohere/north-mini-code:free (1.2 s) and dots-studio/dots-3-note-preview:free (1.2 s) stay steady. If you need a vision‑heavy pipeline, the newly reclaimed poolside/laguna-s-2.1:free (1.6 s) is back in the game, while the heavyweight reasoning models like nvidia/nemotron-3-ultra-550b-a55b (1.0 s) and meta/muse-glimmer-30b (15.0 s) still support tool calls but sit in the slower tier. The churn today is a reminder that even the “stable” models can drop offline in the blink of an eye.

For builders on free tiers, the takeaway is simple: treat the roster as a rotating relay. Sub‑second workers should be your primary lane, with minute‑range options serving as niche fallbacks. The surprise loss of poolside/laguna-s-2.1:free, google/gemma-4-31b-it, and meta/llama-3.2-90b-vision-instruct (now stuck behind auth errors) shows that a model’s tool flag can evaporate without warning. Keep a queue moving, spin through the passing models before the next lock‑out hits, and never assume permanence for any free endpoint.

Tech Talk Tuesday — 2026-10-12 to 2026-10-13

This week’s tool‑call picture is one of modest turnover. Gained tool‑call capability: qwen/qwen3-coder-plus:free, auto:free, deepseek-v4-flash:free, three instances of glm-4.7-flash and aion-labs/aion-3.0. Lost it: apodex/apodex-1.1-mini:free, nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free, poolside/laguna-s-2.1:free, google/gemma-4-31b-it, meta/llama-3.2-90b-vision-instruct and nvidia/ising-calibration-1.5-31b. The churn means anyone relying on a particular tool‑capable model should keep a secondary queue ready; today’s “newcomers” are solid mid‑lane workers, while today’s “exits” remove a few niche vision and reasoning options from the free pool.

This week’s deep dig confirmed a hypothesis: all first‑time passes are still reached via OpenRouter’s proxy, so our measured roster reflects the OpenRouter relay, not the broader free‑tier frontier. No new keyless or direct‑vendor endpoints were observed, and the binary tool flag remains the best proxy for functional capability—though it can be induced with grammar‑constrained decoding if you really want to stretch small models.

To run today’s verified set yourself, clone github.com/fusionpoweredsoftware/anytimelime, set LIME_KEY to an OpenRouter key, and run ./lime.sh. It pulls the day’s roster and rotates across the passing models; your key stays on your own machine.

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
—@cf/meta/llama-4-scout-17b-16e-instructCloudflare Workers AIuntestedDOWN—···
—MiniMax-M2.5-OfficialPooled AIuntestedDOWN—···
0.3Qwen/Qwen2.5-72B-InstructHuggingFace RouteruntestedDOWN—···
3.8qwen/qwen3-coder-plus:freexKiro AIyesFREE3/5···
0.301-ai/yi-largenvidiauntestedDOWN—✗✗✗
0.4DeepSeek-V3.1sambanovauntestedDOWN—✗✗✗
0.3Meta-Llama-3.3-70B-InstructSambaNovauntestedDOWN—✗✗✗
3.8MiniMaxAI/MiniMax-M2.7Gonka BrokeryesFREE3/5✓✓✓
1.2agnes-3.0-flashAgnes AIyesFREE3/5✓✓·
6.1aion-labs/aion-3.0-miniAION LabsyesFREE3/5✓✓·
0.3aisingapore/Gemma-SEA-LION-v4-27B-ITSEA-LIONuntestedDOWN—···
0.3aisingapore/Qwen-SEA-LION-v4.5-27B-ITSEA-LION (AI Singapore)yes (past)DOWN—✗✗✗
3.2auto:freeBazaarLinkyesFREE3/5···
0.4codestral-latestMistral La Plateformeyes (past)DOWN—✗✗✗
1.2cohere/north-mini-code:freeopenrouteryesFREE3/5✓✓✓
1.2cohere/north-mini-code:freeopenrouteryesFREE3/5✓✓✓
0.1deepseek-ai/DeepSeek-V3-0324HuggingFace Inference Routeryes (past)DOWN—✗✗✓
0.5deepseek-ai/DeepSeek-V3.1SiliconFlowyes (past)DOWN—✗✗✗
1.2deepseek-ai/deepseek-v4-flash-0731NVIDIA NIMyes (past)DOWN—✗✗✗
—deepseek-r1:latestZydit AIuntestedDOWN—···
2.0deepseek-v4-flash:freeRouteway AIyesFREE3/5···
1.6dots-studio/dots-3-note-preview:freeopenrouteryesFREE3/5✓✓·
0.4gemini-2.5-flashGoogle AI StudiountestedDOWN—···
0.7glm-4.5-flashZhipu AI (Z.ai)yes (past)DOWN—✗✗·
14.5glm-4.7-flashZ.ai (Zhipu)yesFREE3/5✗✓·
14.5glm-4.7-flashZ.ai (Zhipu)yesFREE3/5✗✓·
14.5glm-4.7-flashZ.ai (Zhipu)yesFREE3/5✗✓·
—glm-5.3-flashZylo APIuntestedDOWN—···
0.2gpt-oss-20bKeylessAIuntestedDOWN—···
0.2gpt-oss:120bOllama ClouduntestedDOWN—···
0.8intern-latestIntern AIyes (past)DOWN—✗✗✗
2.4intern-s2Intern AI (Shanghai AI Lab)yesFREE3/5✓✓·
4.7kilo-auto/freeKilo (Anaconda)yesFREE3/5✓✗·
0.3meganova-ai/manta-flash-1.0MegaNova AIuntestedDOWN—···
0.2meta-llama/Llama-3.3-70B-InstructHugging Face Inferenceyes (past)DOWN—✗✗✓
0.5minimax/minimax-m3:freexKiro AIyes (past)DOWN—✗✗✗
1.7nemotron-3-super-120b-a12b:freeNaga AIyesFREE3/5✓✓·
3.6north-mini-codeTokenReplyyesFREE3/5✓✓✓
1.2openai-fastPollinationsyesFREE3/5✓✓·
0.9openai/gpt-oss-120b:freeFastRouteryesFREE3/5✓✓·
0.5openaiPollinationsyesFREE3/5✓✓✓
1.3qwen/qwen3.5-4b-freeMixlayeryesFREE3/5✓✓·
1.7qwen3-8bFree.aiyesFREE3/5✓✓·
0.5qwen3.8-27bYolo-Autoyes (past)DOWN—✗✗·
60.1tokeness/freeTokenessyes (past)DOWN—✗✓✓
1.1typhoon-v2.5-30b-a3b-instructTyphoon (SCB 10X)yesFREE3/5✓✓·
1.2adept/fuyu-8bnvidiauntestedDOWN—✗✗✗
1.3DeepSeek-V3.2sambanovauntestedDOWN—✗✗✗
0.4MiniMax-M3SambaNovauntestedDOWN—✗✗✗
2.5MiniMax-M2.7-OfficialPooled AIyesFREE3/5✓✓·
0.3deepseek-ai/DeepSeek-V4-Flash-0731Gonka Brokeryes (past)DOWN—✗✗✗
0.3Qwen/Qwen3-Coder-480B-A35B-InstructHuggingFace Routeryes (past)DOWN—✗✗·
3.1aion-labs/aion-3.0AION LabsyesFREE3/5✗✗✗
1.3mistral-largeMistral La PlateformeuntestedDOWN—···
1.6dots-studio/dots-3-note-preview:freeopenrouteryesFREE3/5✓✓·
0.4inclusionAI/Ling-flash-2.0SiliconFlowyes (past)DOWN—✗✗✗
0.8nvidia/nemotron-3-super-120b-a12bNVIDIA NIMyesFREE3/5✓✓✓
0.8intern-latestIntern AIyes (past)DOWN—✗✗✗
0.5openaiPollinationsyesFREE3/5✓✓✓
1.0openai/gpt-oss-20b:freeFastRouteryesFREE3/5✓✓✓
0.3ai21labs/jamba-1.5-large-instructnvidiauntestedDOWN—✗✗✗
0.3gemma-4-31B-itsambanovauntestedDOWN—✗✗✗
0.3gpt-oss-120bSambaNovauntestedDOWN—✗✗·
0.3deepseek-ai/DeepSeek-V4-Flash-0731Gonka Brokeryes (past)DOWN—✗✗✗
0.3google/gemma-4-26b-a4b-it:freeopenrouteryes (past)DOWN—✗✗✗
1.8intern-s2-previewIntern AIyesFREE3/5✓✓·
0.5openaiPollinationsyesFREE3/5✓✓✓
0.2aisingapore/sea-lion-7b-instructnvidiauntestedDOWN—✗✗✗
3.2zai-org/GLM-5.3-FlashGonka BrokeryesFREE3/5✓✓✗
0.3google/gemma-4-31b-it:freeopenrouteryes (past)DOWN—✗✗✗
0.2bigcode/starcoder2-15bnvidiauntestedDOWN—✗✗✗
0.6liquid/lfm-2.5-2.6b:freeopenrouteryesFREE3/5✓✓✓
0.2databricks/dbrx-instructnvidiauntestedDOWN—✗✗✗
0.3nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:freeopenrouteryes (past)DOWN—✓✓✗
0.2deepseek-ai/deepseek-coder-6.7b-instructnvidiauntestedDOWN—✗✗✗
2.4nvidia/nemotron-3-super-120b-a12b:freeopenrouteryesFREE3/5✓✓✓
60.1deepseek-ai/deepseek-v4.1-flashnvidiauntestedDOWN—✗✗✗
11.5nvidia/nemotron-3-ultra-550b-a55b:freeopenrouteryesFREE3/5✓✓✓
0.2google/codegemma-1.1-7bnvidiauntestedDOWN—✗✗✗
0.5nvidia/nemotron-3.5-content-safety:freeopenrouteruntestedDOWN—✗✗✗
0.2google/codegemma-7bnvidiauntestedDOWN—✗✗✗
13.5nvidia/nemotron-3.5-lightning:freeopenrouteryesFREE3/5✓✗✓
0.2google/deplotnvidiauntestedDOWN—✗✗✗
1.7openrouter/freeopenrouteryesFREE3/5✓✓✓
0.5google/diffusiongemma-26b-a4b-itnvidiayesFREE3/5✓✓✓
0.5poolside/laguna-s-2.1:freeopenrouteryes (past)DOWN—✓✓✗
0.4google/gemma-2bnvidiauntestedDOWN—✗✗✗
0.7poolside/laguna-xs-2.1:freeopenrouteryes (past)DOWN—✗✗✗
0.4google/gemma-3-12b-itnvidiauntestedDOWN—✗✗✗
0.6google/gemma-3-4b-itnvidiauntestedDOWN—✗✗✗
60.1google/gemma-4-31b-itnvidiayes (past)DOWN—✓✗✓
0.2google/recurrentgemma-2bnvidiauntestedDOWN—✗✗✗
0.2ibm/granite-3.0-3b-a800m-instructnvidiauntestedDOWN—✗✗✗
0.2ibm/granite-3.0-8b-instructnvidiauntestedDOWN—✗✗✗
0.2ibm/granite-34b-code-instructnvidiauntestedDOWN—✗✗✗
0.2ibm/granite-8b-code-instructnvidiauntestedDOWN—✗✗✗
0.2meta/codellama-70bnvidiauntestedDOWN—✗✗✗
0.8meta/llama-3.2-11b-vision-instructnvidiayesFREE3/5✓✓✓
60.1meta/llama-3.2-90b-vision-instructnvidiayes (past)DOWN—✓✗✓
30.3meta/llama2-70bnvidiauntestedDOWN—✗✗✗
33.8meta/muse-glimmer-30bnvidiayesFREE3/5✓✓✓
0.2microsoft/kosmos-2nvidiauntestedDOWN—✗✗✗
30.2microsoft/phi-3-vision-128k-instructnvidiauntestedDOWN—✗✗✗
0.2microsoft/phi-3.5-moe-instructnvidiauntestedDOWN—✗✗✗
0.2mistralai/codestral-22b-instruct-v0.1nvidiauntestedDOWN—✗✗✗
30.3mistralai/mistral-7b-instruct-v0.3nvidiauntestedDOWN—✗✗✗
30.2mistralai/mistral-large-2-instructnvidiauntestedDOWN—✗✗✗
0.2mistralai/mistral-largenvidiauntestedDOWN—✗✗✗
0.3mistralai/mixtral-8x22b-v0.1nvidiauntestedDOWN—✗✗✗
0.2moonshotai/kimi-k2.6nvidiauntestedDOWN—✗✗✗
57.8moonshotai/kimi-k3nvidiayesFREE3/5✓✗✗
0.3nv-mistralai/mistral-nemo-12b-instructnvidiauntestedDOWN—✗✗✗
0.3nvidia/ai-synthetic-video-detectornvidiauntestedDOWN—✗✗✗
0.2nvidia/cosmos-reason2-8bnvidiauntestedDOWN—✗✗✗
0.9nvidia/ising-calibration-1.5-31bnvidiatext-onlyFREE—✓✓✓
30.2nvidia/llama-3.1-nemotron-51b-instructnvidiauntestedDOWN—✗✗✗
30.2nvidia/llama-3.1-nemotron-70b-instructnvidiauntestedDOWN—✗✗✗
30.2nvidia/llama-3.1-nemotron-ultra-253b-v1nvidiauntestedDOWN—✗✗✗
30.2nvidia/llama3-chatqa-1.5-70bnvidiauntestedDOWN—✗✗✗
0.2nvidia/mistral-nemo-minitron-8b-8k-instructnvidiauntestedDOWN—✗✗✗
2.1nvidia/nemotron-3-nano-omni-30b-a3b-reasoningnvidiayesFREE3/5✓✓✓
2.2nvidia/nemotron-3-ultra-550b-a55bnvidiayesFREE3/5✓✓✓
0.2nvidia/nemotron-3.5-content-safetynvidiauntestedDOWN—✗✗✗
1.8nvidia/nemotron-3.5-lightning-30b-a3bnvidiayesFREE3/5✓✓✓
0.2nvidia/nemotron-4-340b-instructnvidiauntestedDOWN—✗✗✗
0.2nvidia/nemotron-4-340b-rewardnvidiauntestedDOWN—✗✗✗
0.2nvidia/nemotron-nano-3-30b-a3bnvidiauntestedDOWN—✗✗✗
0.3nvidia/nemotron-parse-2.0nvidiauntestedDOWN—✗✗✗
0.3nvidia/nemotron-parsenvidiauntestedDOWN—✗✗✗
0.2nvidia/neva-22bnvidiauntestedDOWN—✗✗✗
0.2nvidia/nvclipnvidiauntestedDOWN—✗✗✗
60.1nvidia/riva-translate-4b-instruct-v2nvidiauntestedDOWN—✗✗✗
0.2nvidia/riva-translate-4b-instructnvidiauntestedDOWN—✗✗✗
0.2nvidia/vilanvidiauntestedDOWN—✗✗✗

43 made a tool call, 1 answered in text only, of 134 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.