anytimelime — the free AI field report

2026-09-26 · generated by a free model interviewing free models · anytimelime, formerly Project42 · get the engine →

What's free in AI right now — 2026-09-26

The morning’s headline is a quiet but sweeping lock on SambaNova’s free lane. Six recent models that once answered tools without a bill—DeepSeek‑V3.1, DeepSeek‑V3.2, Meta‑Llama‑3.3‑70B‑Instruct, MiniMax‑M2.7, MiniMax‑M3, gemma‑4‑31B‑it and openai‑gpt‑oss‑120b—are now behind a “payment method required” screen. The probe never reached the model itself, but for anyone relying on OpenRouter’s `:free` tier the gate is effectively closed. On the flip side, three models that had been silent are back in the rotation: tokeness/free, zai-org/GLM‑5.3‑Flash and nvidia/nemotron‑3‑super‑120b‑a12b all logged clean tool calls again. The churn is the story—paywalls tightening on one side, old hands back on the other. Running the free tier today feels like a staggered lineup. The newly resurgent tokens dominate the latency charts: tokeness/free answered in 9.2 s, zai-org/GLM‑5.3‑Flash in 4.0 s and nvidia/nemotron‑3‑super‑120b‑a12b in a brisk 0.8 s. If you need a reliable workhorse, the usual suspects hold steady: MiniMax‑M2.7‑Official (2.9 s), aion‑labs/aion‑3.0 (3.6 s), agnes‑3.0‑flash (1.1 s) and openai‑fast (1.0 s). OpenRouter’s openai‑gpt‑oss‑120b:free and cohere/north‑mini‑code:free keep tool use humming at sub‑second speeds, while nvidia/nemotron‑3‑nano‑omni‑30b‑a3b‑reasoning:free adds a 3.1 s burst of reasoning. A few veterans have slipped: nvidia/nemotron‑3‑5‑lightning:free now times out after 60 s and poolside/laguna‑s‑2.1:free returns a provider error. The roster is a quick‑draw of comebacks and lockouts. For builders on free tiers, today’s shuffle is a reminder that rotation isn’t a convenience but a survival tactic. A model that answered tools yesterday can be silent today, and a paywall shift can strip a “free” label without warning. The latency spread—from sub‑second openai‑fast to double‑digit tokeness—means scripts must be flexible enough to pick the cheapest, fastest option before a gate closes. The only constant is churn, and any dependable stack must re‑test before it’s trusted. Clone the engine from github.com/fusionpoweredsoftware/anytimelime, set `LIME_KEY` to your OpenRouter key, and run `./lime.sh`. It pulls today’s verified roster, skips the empty‑response dead ends, and rotates across the passing models so no single free tier carries all your traffic. Your key never leaves the reader’s machine.

Today's passing roster

openai — Pollinations · 0.3s · agentic 3/5
openai — Pollinations · 0.3s · agentic 3/5
typhoon-v2.5-30b-a3b-instruct — Typhoon (SCB 10X) · 0.7s · agentic 3/5
inclusionai/ling-3.0-flash-fin:free — openrouter · 0.8s · agentic 3/5
google/diffusiongemma-26b-a4b-it — nvidia · 0.8s · agentic 3/5
openrouter/free — openrouter · 0.8s · agentic 3/5
nvidia/nemotron-3-super-120b-a12b — nvidia · 0.8s · agentic 3/5
openai/gpt-oss-120b:free — FastRouter · 0.9s · agentic 3/5
nvidia/nemotron-3-ultra-550b-a55b:free — openrouter · 0.9s · agentic 3/5
openai-fast — Pollinations · 1.0s · agentic 3/5
nvidia/nemotron-3-ultra-550b-a55b — nvidia · 1.0s · agentic 3/5
agnes-3.0-flash — Agnes AI · 1.1s · agentic 3/5
deepseek-ai/DeepSeek-V3-0324 — HuggingFace Inference Router · 1.1s · agentic 3/5
stealth/space-bunny-alpha — openrouter · 1.1s · agentic 3/5
north-mini-code — TokenReply · 1.3s · agentic 3/5
inclusionai/ling-3.0-flash-sante:free — openrouter · 1.4s · agentic 3/5
qwen/qwen3.5-4b-free — Mixlayer · 1.5s · agentic 3/5
qwen3-8b — Free.ai · 1.5s · agentic 3/5
dots-studio/dots-3-note-preview:free — openrouter · 1.6s · agentic 3/5
liquid/lfm-2.5-2.6b:free — openrouter · 1.6s · agentic 3/5
MiniMaxAI/MiniMax-M2.7 — Gonka Broker · 2.1s · agentic 3/5
cohere/north-mini-code:free — openrouter · 2.1s · agentic 3/5
nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free — openrouter · 2.1s · agentic 3/5
nvidia/nemotron-3-super-120b-a12b:free — openrouter · 2.1s · agentic 3/5
nemotron-3-super-120b-a12b:free — Naga AI · 2.3s · agentic 3/5
meta/muse-glimmer-30b — nvidia · 2.8s · agentic 3/5
MiniMax-M2.7-Official — Pooled AI · 2.9s · agentic 3/5
nvidia/nemotron-3-nano-omni-30b-a3b-reasoning — nvidia · 2.9s · agentic 3/5
google/gemma-4-31b-it — nvidia · 3.1s · agentic 3/5
aion-labs/aion-3.0 — AION Labs · 3.6s · agentic 3/5
zai-org/GLM-5.3-Flash — Gonka Broker · 4.0s · agentic 3/5
moonshotai/kimi-k3 — nvidia · 5.1s · agentic 3/5
tokeness/free — Tokeness · 9.2s · agentic 3/5
nvidia/nemotron-3.5-lightning-30b-a3b — nvidia · 9.6s · agentic 3/5
meta/llama-3.2-11b-vision-instruct — nvidia · 12.4s · agentic 3/5

The measured table

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

latency	model	vendor	tool call	free now	agentic	yesterday	3 days ago	a week ago
0.2	`01-ai/yi-large`	nvidia	untested	DOWN	—	✗	✗	✗
0.3	`DeepSeek-V3.1`	sambanova	untested	CARD	—	✗	✗	✗
2.9	`MiniMax-M2.7-Official`	Pooled AI	yes	FREE	3/5	✓	✓	·
2.1	`MiniMaxAI/MiniMax-M2.7`	Gonka Broker	yes	FREE	3/5	✓	✓	·
1.1	`agnes-3.0-flash`	Agnes AI	yes	FREE	3/5	✓	✓	·
3.6	`aion-labs/aion-3.0`	AION Labs	yes	FREE	3/5	✓	✓	·
0.4	`aisingapore/Qwen-SEA-LION-v4.5-27B-IT`	SEA-LION (AI Singapore)	yes (past)	DOWN	—	✗	✗	·
0.3	`codestral-latest`	Mistral	yes (past)	DOWN	—	✗	✗	·
2.1	`cohere/north-mini-code:free`	openrouter	yes	FREE	3/5	✓	✓	✓
1.1	`deepseek-ai/DeepSeek-V3-0324`	HuggingFace Inference Router	yes	FREE	3/5	✓	✓	·
0.9	`deepseek-ai/DeepSeek-V3.1`	SiliconFlow	yes (past)	DOWN	—	✗	·	·
0.1	`deepseek-ai/deepseek-v4-flash-0731`	NVIDIA NIM	yes (past)	DOWN	—	✗	✗	✓
0.8	`glm-4.5-flash`	Zhipu AI (Z.ai)	yes (past)	DOWN	—	✗	✓	·
0.7	`glm-4.7-flash`	Zhipu AI (Z.ai)	yes (past)	DOWN	—	✗	·	·
1.6	`intern-latest`	Intern AI	yes (past)	DOWN	—	✗	✓	·
0.4	`minimax/minimax-m3:free`	xKiro AI	yes (past)	DOWN	—	✗	✓	·
2.3	`nemotron-3-super-120b-a12b:free`	Naga AI	yes	FREE	3/5	✓	✓	·
1.3	`north-mini-code`	TokenReply	yes	FREE	3/5	✓	✓	·
1.0	`openai-fast`	Pollinations	yes	FREE	3/5	✓	✓	·
0.9	`openai/gpt-oss-120b:free`	FastRouter	yes	FREE	3/5	✓	✓	·
0.3	`openai`	Pollinations	yes	FREE	3/5	✓	✓	·
1.5	`qwen/qwen3.5-4b-free`	Mixlayer	yes	FREE	3/5	✓	✓	·
1.5	`qwen3-8b`	Free.ai	yes	FREE	3/5	✓	✓	·
0.2	`qwen3.8-27b`	Yolo-Auto	yes (past)	DOWN	—	✗	✓	·
9.2	`tokeness/free`	Tokeness	yes	FREE	3/5	✗	✓	·
0.7	`typhoon-v2.5-30b-a3b-instruct`	Typhoon (SCB 10X)	yes	FREE	3/5	✓	·	·
0.1	`adept/fuyu-8b`	nvidia	untested	DOWN	—	✗	✗	✗
0.3	`DeepSeek-V3.2`	sambanova	untested	CARD	—	✗	✗	✗
0.5	`deepseek-ai/DeepSeek-V4-Flash-0731`	Gonka Broker	yes (past)	DOWN	—	✗	✗	·
1.6	`dots-studio/dots-3-note-preview:free`	openrouter	yes	FREE	3/5	✓	✓	✓
0.2	`inclusionAI/Ling-flash-2.0`	SiliconFlow	yes (past)	DOWN	—	✗	·	·
0.7	`glm-4.7-flash`	Zhipu AI (Z.ai)	yes (past)	DOWN	—	✗	·	·
0.3	`openai`	Pollinations	yes	FREE	3/5	✓	✓	·
0.1	`ai21labs/jamba-1.5-large-instruct`	nvidia	untested	DOWN	—	✗	✗	✗
0.3	`Meta-Llama-3.3-70B-Instruct`	sambanova	untested	CARD	—	✗	✗	✗
4.0	`zai-org/GLM-5.3-Flash`	Gonka Broker	yes	FREE	3/5	✗	✓	·
0.2	`google/gemma-4-26b-a4b-it:free`	openrouter	yes (past)	DOWN	—	✗	✗	✗
0.2	`aisingapore/sea-lion-7b-instruct`	nvidia	untested	DOWN	—	✗	✗	✗
0.3	`MiniMax-M2.7`	sambanova	untested	CARD	—	✗	✗	✗
0.3	`google/gemma-4-31b-it:free`	openrouter	yes (past)	DOWN	—	✗	✗	✓
0.1	`bigcode/starcoder2-15b`	nvidia	untested	DOWN	—	✗	✗	✗
0.3	`MiniMax-M3`	sambanova	untested	CARD	—	✗	✗	✗
0.8	`inclusionai/ling-3.0-flash-fin:free`	openrouter	yes	FREE	3/5	✓	✓	✓
0.1	`databricks/dbrx-instruct`	nvidia	untested	DOWN	—	✗	✗	✗
0.3	`gemma-4-31B-it`	sambanova	untested	CARD	—	✗	✗	✗
1.4	`inclusionai/ling-3.0-flash-sante:free`	openrouter	yes	FREE	3/5	✓	✓	·
0.1	`deepseek-ai/deepseek-coder-6.7b-instruct`	nvidia	untested	DOWN	—	✗	✗	✗
0.3	`gpt-oss-120b`	sambanova	untested	CARD	—	✗	✗	✗
1.6	`liquid/lfm-2.5-2.6b:free`	openrouter	yes	FREE	3/5	✓	✓	✓
60.1	`deepseek-ai/deepseek-v4.1-flash`	nvidia	untested	DOWN	—	✗	✗	·
2.1	`nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free`	openrouter	yes	FREE	3/5	✓	✓	·
0.1	`google/codegemma-1.1-7b`	nvidia	untested	DOWN	—	✗	✗	·
2.1	`nvidia/nemotron-3-super-120b-a12b:free`	openrouter	yes	FREE	3/5	✓	✓	·
0.1	`google/codegemma-7b`	nvidia	untested	DOWN	—	✗	✗	·
0.9	`nvidia/nemotron-3-ultra-550b-a55b:free`	openrouter	yes	FREE	3/5	✓	✓	·
0.2	`google/deplot`	nvidia	untested	DOWN	—	✗	✗	·
0.2	`nvidia/nemotron-3.5-content-safety:free`	openrouter	untested	DOWN	—	✗	✗	·
0.8	`google/diffusiongemma-26b-a4b-it`	nvidia	yes	FREE	3/5	✓	✓	·
60.1	`nvidia/nemotron-3.5-lightning:free`	openrouter	yes (past)	DOWN	—	✓	✓	·
0.1	`google/gemma-2b`	nvidia	untested	DOWN	—	✗	✗	·
0.8	`openrouter/free`	openrouter	yes	FREE	3/5	✓	✓	·
0.2	`google/gemma-3-12b-it`	nvidia	untested	DOWN	—	✗	✗	·
0.2	`poolside/laguna-s-2.1:free`	openrouter	yes (past)	DOWN	—	✓	✓	·
0.1	`google/gemma-3-4b-it`	nvidia	untested	DOWN	—	✗	✗	·
0.3	`poolside/laguna-xs-2.1:free`	openrouter	yes (past)	DOWN	—	✗	✗	·
3.1	`google/gemma-4-31b-it`	nvidia	yes	FREE	3/5	✓	✓	·
0.2	`qwen/qwen3.8-27b:free`	openrouter	yes (past)	DOWN	—	✗	✗	·
0.1	`google/recurrentgemma-2b`	nvidia	untested	DOWN	—	✗	✗	·
1.1	`stealth/space-bunny-alpha`	openrouter	yes	FREE	3/5	✓	✓	·
0.1	`ibm/granite-3.0-3b-a800m-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`ibm/granite-3.0-8b-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`ibm/granite-34b-code-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`ibm/granite-8b-code-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`meta/codellama-70b`	nvidia	untested	DOWN	—	✗	✗	·
12.4	`meta/llama-3.2-11b-vision-instruct`	nvidia	yes	FREE	3/5	✓	✓	·
60.1	`meta/llama-3.2-90b-vision-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`meta/llama2-70b`	nvidia	untested	DOWN	—	✗	✗	·
2.8	`meta/muse-glimmer-30b`	nvidia	yes	FREE	3/5	✓	✓	·
0.1	`microsoft/kosmos-2`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`microsoft/phi-3-vision-128k-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`microsoft/phi-3.5-moe-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`mistralai/codestral-22b-instruct-v0.1`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`mistralai/mistral-7b-instruct-v0.3`	nvidia	untested	DOWN	—	✗	✗	·
0.2	`mistralai/mistral-large-2-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`mistralai/mistral-large`	nvidia	untested	DOWN	—	✗	✗	·
60.1	`mistralai/mistral-nemotron`	nvidia	yes (past)	DOWN	—	✗	✗	·
0.1	`mistralai/mixtral-8x22b-v0.1`	nvidia	untested	DOWN	—	✗	✗	·
0.2	`moonshotai/kimi-k2.6`	nvidia	untested	DOWN	—	✗	✗	·
5.1	`moonshotai/kimi-k3`	nvidia	yes	FREE	3/5	✓	✗	·
0.1	`nv-mistralai/mistral-nemo-12b-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.2	`nvidia/ai-synthetic-video-detector`	nvidia	untested	DOWN	—	✗	✗	·
0.2	`nvidia/cosmos-reason2-8b`	nvidia	untested	DOWN	—	✗	✗	·
1.6	`nvidia/ising-calibration-1.5-31b`	nvidia	text-only	FREE	—	✓	✗	·
0.2	`nvidia/llama-3.1-nemotron-51b-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`nvidia/llama-3.1-nemotron-70b-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`nvidia/llama-3.1-nemotron-ultra-253b-v1`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`nvidia/llama3-chatqa-1.5-70b`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`nvidia/mistral-nemo-minitron-8b-8k-instruct`	nvidia	untested	DOWN	—	✗	✗	·
2.9	`nvidia/nemotron-3-nano-omni-30b-a3b-reasoning`	nvidia	yes	FREE	3/5	✓	✓	·
0.8	`nvidia/nemotron-3-super-120b-a12b`	nvidia	yes	FREE	3/5	✗	✓	·
1.0	`nvidia/nemotron-3-ultra-550b-a55b`	nvidia	yes	FREE	3/5	✓	✓	·
0.2	`nvidia/nemotron-3.5-content-safety`	nvidia	untested	DOWN	—	✗	✗	·
9.6	`nvidia/nemotron-3.5-lightning-30b-a3b`	nvidia	yes	FREE	3/5	✓	✓	·
0.1	`nvidia/nemotron-4-340b-instruct`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`nvidia/nemotron-4-340b-reward`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`nvidia/nemotron-nano-3-30b-a3b`	nvidia	untested	DOWN	—	✗	✗	·
0.2	`nvidia/nemotron-parse-2.0`	nvidia	untested	DOWN	—	✗	✗	·
0.2	`nvidia/nemotron-parse`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`nvidia/neva-22b`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`nvidia/nvclip`	nvidia	untested	DOWN	—	✗	✗	·
0.4	`nvidia/riva-translate-4b-instruct-v1.1`	nvidia	untested	DOWN	—	✗	✗	·
0.1	`nvidia/riva-translate-4b-instruct`	nvidia	untested	DOWN	—	✗	✗	·

35 made a tool call, 1 answered in text only, of 112 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.