anytimelime — the free AI field report

2026-10-06 · generated by a free model interviewing free models · anytimelime, formerly Project42 · get the engine →

What's free in AI right now — 2026-10-06

Today’s headline: tokeness/free makes a comeback, MiniMax drifts away

The most interesting shuffle today is tokeness/free flashing back to life with a tool‑yes flag after weeks of silence. It now answers at a sluggish 59 seconds but does pass the tool call, a reminder that even slow free tiers can be a lifeline for batch jobs. In the same churn, two of NVIDIA’s Nemotron models—nvidia/nemotron-3-ultra-550b-a55b:free (4.9 s) and nvidia/nemotron-3-nano-omni-30b-a3b-reasoning (3.4 s)—graduated to tool‑capable status, expanding the free‑tool pool. The surprise dropout is MiniMax-M2.7-Official (2.5 s), which lost its tool flag after a brief run, leaving its sibling MiniMaxAI/MiniMax-M2.7 as the sole MiniMax worker still serving functions.

What it feels like to fire up the free tier right now

If you need a tool‑ready model that doesn’t cost a penny, start with the ultra‑fast Pollinations openai endpoint (0.3 s) – it still talks tools without a key. The Hugging Face‑hosted meta-llama/Llama-3.3-70B-Instruct (0.6 s) remains a solid workhorse for reasoning tasks. NVIDIA’s newest tool‑enabled Nemotrons are worth a spin: nvidia/nemotron-3-ultra-550b-a55b:free for heavy‑weight prompts and the leaner nvidia/nemotron-3-nano-omni-30b-a3b-reasoning for quick decisions. If you can tolerate latency in the minute‑range, tokeness/free is the only source that now offers tool calls again, making it a niche fallback for non‑urgent batch pipelines. Vision models like meta/llama-3.2-90b-vision-instruct (32.3 s) and meta/llama-3.2-11b-vision-instruct (0.8 s) still hit the tool mark, giving you a rare glimpse of free multimodal capability.

What this churn means for free‑tier builders

Today's roster proves that the free pool is a revolving door. Yesterday’s reliable tool‑callers can vanish without warning, while silent models like tokeness/free can resurface with a new tool flag. For production scripts, treat latency as the first guardrail: sub‑second workers (openai, meta-llama/Llama-3.3-70B-Instruct) should be the primary lane, while longer‑wait options become secondary fallbacks. Keep a rotation queue in motion – spin through the passing models before the next lock‑out hits, and never assume a “free” label is permanent. The only steady bet is that the free tier will keep you on your toes.

To run the day’s verified set yourself, clone github.com/fusionpoweredsoftware/anytimelime, set LIME_KEY to an OpenRouter key, and execute ./lime.sh. It fetches today’s roster and rotates across the passing models; your key never leaves your 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
—Meta-Llama-3.1-8B-InstructSambaNova ClouduntestedDOWN—···
—Qwen/Qwen2.5-7B-InstructSiliconFlowuntestedDOWN—···
—Qwen2.5-3BZenith-AIuntestedDOWN—···
2.3Qwen3.6-27BOVHcloud AI EndpointsuntestedDOWN—···
—glm-5.3Zylo APIuntestedDOWN—···
—kimi-k2.5LogfareuntestedDOWN—···
—qwen/qwen3.8-max:freexKiro AIuntestedDOWN—···
0.301-ai/yi-largenvidiauntestedDOWN—✗✗✗
0.2@cf/meta/llama-3.3-70b-instruct-fp8-fastCloudflare Workers AIuntestedDOWN—···
0.6@cf/zai-org/glm-5.2PersorAIuntestedDOWN—··✗
0.3DeepSeek-V3.1sambanovauntestedDOWN—✗✗✗
—MiniMax-M2.7-OfficialPooled AIyes (past)DOWN—✓✓✓
7.3MiniMaxAI/MiniMax-M2.7Gonka BrokeryesFREE3/5✓✓✓
—Qwen/Qwen3-Coder-480B-A35B-InstructHuggingFace Routeryes (past)DOWN—✗✓✓
—Qwen/Qwen3-Coder-480B-A35B-InstructHuggingFace Routeryes (past)DOWN—✗✓✓
—agnes-3.0-flashAgnes AIyes (past)DOWN—✓✓✓
0.3aion-labs/aion-3.0AION Labsyes (past)DOWN—✗✓✓
0.6aisingapore/Qwen-SEA-LION-v4.5-27B-ITSEA-LIONyes (past)DOWN—✗✗✗
1.1apodex/apodex-1.1-mini:freeopenrouteryesFREE3/5✓✓·
0.3codestral-latestMistralyes (past)DOWN—✗✗✗
0.8deepseek-ai/DeepSeek-V3-0324HuggingFace Inference RouteryesFREE3/5✓✓✓
0.1deepseek-ai/deepseek-v4-flash-0731NVIDIA NIMyes (past)DOWN—✗✗✗
—deepseek-r1:latestZydit AIuntestedDOWN—···
—dots-studio/dots-3-note-preview:freeopenrouteryes (past)DOWN—✓✓✓
—glm-4.5-flashZhipu AIyes (past)DOWN—✗✗✗
—glm-4.7-flashZhipu AI (Z.ai)yes (past)DOWN—✓✓✓
—glm-4.7-flashZhipu AI (Z.ai)yes (past)DOWN—✓✓✓
—glm-4.7-flashZhipu AI (Z.ai)yes (past)DOWN—✓✓✓
—gpt-oss-120bsambanovauntestedDOWN—✗✗✗
—gpt-oss:120bOllama ClouduntestedDOWN—···
1.1intern-latestIntern AIyes (past)DOWN—✗✗✗
0.6meta-llama/Llama-3.3-70B-InstructHugging Face InferenceyesFREE3/5✓✓✓
—minimax-m2.7:freeRouteway AIuntestedDOWN—···
0.8minimax/minimax-m3:freexKiro AIyes (past)DOWN—✗✗✗
0.7mistralai/Mistral-Small-3.2-24B-Instruct-2506MegaNova AIuntestedDOWN—···
—nemotron-3-super-120b-a12b:freeNaga AIyes (past)DOWN—✓✓✓
1.7north-mini-codeTokenReplyyesFREE3/5✓✓✓
0.1openai-fastKeylessAIuntestedDOWN—···
0.1openai-fastKeylessAIuntestedDOWN—···
—openai/gpt-oss-120b:freeFastRouteryes (past)DOWN—✓✓✓
0.3openaiPollinationsyesFREE3/5✓✓✓
—orcarouter/freeOrcarouteruntestedDOWN—···
—qwen/qwen3.5-4b-freeMixlayeryes (past)DOWN—✓✓✓
—qwen3-8bFree.aiyes (past)DOWN—✓✓✓
—qwen3.8-27bYolo-Autoyes (past)DOWN—✗✗✗
59.3tokeness/freeTokenessyesFREE3/5✗✓✗
—typhoon-v2.5-30b-a3b-instructTyphoon (SCB 10X)yes (past)DOWN—✓✓✓
0.1adept/fuyu-8bnvidiauntestedDOWN—✗✗✗
0.3DeepSeek-V3.2sambanovauntestedDOWN—✗✗✗
0.5deepseek-ai/DeepSeek-V4-Flash-0731Gonka Brokeryes (past)DOWN—✗✗✗
0.8deepseek-ai/DeepSeek-V3.1SiliconFlowyes (past)DOWN—✗✗✗
1.1cohere/north-mini-code:freeopenrouteryesFREE3/5✓✓✓
—intern-s2-previewIntern AIyes (past)DOWN—✓✓✓
0.3openaiPollinationsyesFREE3/5✓✓✓
1.3openai/gpt-oss-20b:freeFastRouteryesFREE3/5✓✓✓
0.2ai21labs/jamba-1.5-large-instructnvidiauntestedDOWN—✗✗✗
0.3Meta-Llama-3.3-70B-InstructsambanovauntestedDOWN—✗✗✗
0.5zai-org/GLM-5.3-FlashGonka Brokeryes (past)DOWN—✓✓✓
0.2inclusionAI/Ling-flash-2.0SiliconFlowyes (past)DOWN—✗✗✗
—dots-studio/dots-3-note-preview:freeopenrouteryes (past)DOWN—✓✓✓
0.1aisingapore/sea-lion-7b-instructnvidiauntestedDOWN—✗✗✗
0.3MiniMax-M3sambanovauntestedDOWN—✗✗✗
0.3google/gemma-4-26b-a4b-it:freeopenrouteryes (past)DOWN—✗✗✗
0.1bigcode/starcoder2-15bnvidiauntestedDOWN—✗✗✗
0.3gemma-4-31B-itsambanovauntestedDOWN—✗✗✗
0.2google/gemma-4-31b-it:freeopenrouteryes (past)DOWN—✗✗✗
0.1databricks/dbrx-instructnvidiauntestedDOWN—✗✗✗
—gpt-oss-120bsambanovauntestedDOWN—✗✗✗
1.1inclusionai/ling-3.0-flash-sante:freeopenrouteryesFREE3/5✓✓✓
0.1deepseek-ai/deepseek-coder-6.7b-instructnvidiauntestedDOWN—✗✗✗
0.2inclusionai/ling-3.1-flashopenrouteruntestedCARD—✗✗·
60.1deepseek-ai/deepseek-v4.1-flashnvidiauntestedDOWN—✗✗✗
0.9liquid/lfm-2.5-2.6b:freeopenrouteryesFREE3/5✓✓✓
0.2google/codegemma-1.1-7bnvidiauntestedDOWN—✗✗✗
0.3nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:freeopenrouteryes (past)DOWN—✗✓✓
0.2google/codegemma-7bnvidiauntestedDOWN—✗✗✗
1.0nvidia/nemotron-3-super-120b-a12b:freeopenrouteryesFREE3/5✓✓✗
0.2google/deplotnvidiauntestedDOWN—✗✗✗
4.9nvidia/nemotron-3-ultra-550b-a55b:freeopenrouteryesFREE3/5✗✓✓
0.5google/diffusiongemma-26b-a4b-itnvidiayesFREE3/5✓✗✓
0.1nvidia/nemotron-3.5-content-safety:freeopenrouteruntestedDOWN—✗✗✗
0.1google/gemma-2bnvidiauntestedDOWN—✗✗✗
1.5nvidia/nemotron-3.5-lightning:freeopenrouteryesFREE3/5✓✓✓
0.1google/gemma-3-12b-itnvidiauntestedDOWN—✗✗✗
1.4openrouter/freeopenrouteryesFREE3/5✓✓✓
0.1google/gemma-3-4b-itnvidiauntestedDOWN—✗✗✗
0.3poolside/laguna-s-2.1:freeopenrouteryes (past)DOWN—✗✓✗
24.1google/gemma-4-31b-itnvidiayesFREE3/5✓✓✓
0.3poolside/laguna-xs-2.1:freeopenrouteryes (past)DOWN—✗✗✗
0.1google/recurrentgemma-2bnvidiauntestedDOWN—✗✗✗
0.1ibm/granite-3.0-3b-a800m-instructnvidiauntestedDOWN—✗✗✗
0.1ibm/granite-3.0-8b-instructnvidiauntestedDOWN—✗✗✗
0.1ibm/granite-34b-code-instructnvidiauntestedDOWN—✗✗✗
0.1ibm/granite-8b-code-instructnvidiauntestedDOWN—✗✗✗
0.1meta/codellama-70bnvidiauntestedDOWN—✗✗✗
0.8meta/llama-3.2-11b-vision-instructnvidiayesFREE3/5✓✓✓
32.3meta/llama-3.2-90b-vision-instructnvidiayesFREE3/5✓✓✓
0.1meta/llama2-70bnvidiauntestedDOWN—✗✗✗
3.8meta/muse-glimmer-30bnvidiayesFREE3/5✓✓✓
0.1microsoft/kosmos-2nvidiauntestedDOWN—✗✗✗
0.1microsoft/phi-3-vision-128k-instructnvidiauntestedDOWN—✗✗✗
0.1microsoft/phi-3.5-moe-instructnvidiauntestedDOWN—✗✗✗
0.2mistralai/codestral-22b-instruct-v0.1nvidiauntestedDOWN—✗✗✗
0.1mistralai/mistral-7b-instruct-v0.3nvidiauntestedDOWN—✗✗✗
0.1mistralai/mistral-large-2-instructnvidiauntestedDOWN—✗✗✗
0.1mistralai/mistral-largenvidiauntestedDOWN—✗✗✗
0.1mistralai/mixtral-8x22b-v0.1nvidiauntestedDOWN—✗✗✗
0.1moonshotai/kimi-k2.6nvidiauntestedDOWN—✗✗✗
60.1moonshotai/kimi-k3nvidiayes (past)DOWN—✓✓✗
0.1nv-mistralai/mistral-nemo-12b-instructnvidiauntestedDOWN—✗✗✗
0.2nvidia/ai-synthetic-video-detectornvidiauntestedDOWN—✗✗✗
0.1nvidia/cosmos-reason2-8bnvidiauntestedDOWN—✗✗✗
0.6nvidia/ising-calibration-1.5-31bnvidiatext-onlyFREE—✓✓✓
0.1nvidia/llama-3.1-nemotron-51b-instructnvidiauntestedDOWN—✗✗✗
0.2nvidia/llama-3.1-nemotron-70b-instructnvidiauntestedDOWN—✗✗✗
0.2nvidia/llama-3.1-nemotron-ultra-253b-v1nvidiauntestedDOWN—✗✗✗
0.1nvidia/llama3-chatqa-1.5-70bnvidiauntestedDOWN—✗✗✗
0.2nvidia/mistral-nemo-minitron-8b-8k-instructnvidiauntestedDOWN—✗✗✗
3.4nvidia/nemotron-3-nano-omni-30b-a3b-reasoningnvidiayesFREE3/5✗✗✓
0.9nvidia/nemotron-3-super-120b-a12bnvidiayesFREE3/5✓✓✓
2.7nvidia/nemotron-3-ultra-550b-a55bnvidiayesFREE3/5✓✓✓
0.2nvidia/nemotron-3.5-content-safetynvidiauntestedDOWN—✗✗✗
4.4nvidia/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—✗✗✗
0.2nvidia/nemotron-parse-2.0nvidiauntestedDOWN—✗✗✗
0.2nvidia/nemotron-parsenvidiauntestedDOWN—✗✗✗
0.1nvidia/neva-22bnvidiauntestedDOWN—✗✗✗
0.1nvidia/nvclipnvidiauntestedDOWN—✗✗✗
14.6nvidia/riva-translate-4b-instruct-v2nvidiauntestedDOWN—✗✗✗
0.1nvidia/riva-translate-4b-instructnvidiauntestedDOWN—✗✗✗
0.2nvidia/vilanvidiauntestedDOWN—✗··

29 made a tool call, 1 answered in text only, of 133 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.