[{"data":1,"prerenderedAt":740},["ShallowReactive",2],{"blog-post-chatgpt-codex-shared-usage-limit":3,"related-posts-chatgpt-codex-shared-usage-limit":718},{"id":4,"title":5,"author":6,"body":11,"category":695,"date":696,"description":697,"extension":698,"featured":699,"hideToc":699,"image":700,"imageAlt":701,"imageHeight":701,"imageWidth":701,"itemList":701,"lang":701,"meta":702,"navigation":703,"noindex":699,"path":704,"readingTime":705,"redirected":699,"relatedSlugs":701,"seo":706,"seoTitle":707,"stem":708,"subtitle":701,"tags":709,"updatedDate":696,"__hash__":717},"blog/blog/chatgpt-codex-shared-usage-limit.md","ChatGPT and Codex Shared Usage Limit: What Changes in December",{"name":7,"role":8,"avatar":9,"bio":10},"Shabnam Katoch","Growth Head","/img/avatars/shabnam-profile.jpeg","Shabnam leads growth at BetterClaw and has spent the past three years building organic distribution for AI-native products. She writes about no-code agent builders, content strategy, and what it actually takes to grow without an ads budget.",{"type":12,"value":13,"toc":670},"minimark",[14,18,26,29,32,38,41,46,49,52,55,62,127,130,136,139,143,146,159,166,169,172,175,179,182,185,263,269,275,278,283,287,290,307,332,339,342,348,351,355,358,364,369,372,381,386,389,397,400,404,411,414,418,421,424,428,431,438,442,445,452,456,459,462,465,468,476,485,489,492,495,556,562,565,568,576,580,583,586,589,594,613,617,621,624,628,631,635,641,645,648,652,655,659,666],[15,16,17],"p",{},"Which plans are affected, what counts against the shared pool, and how to stop Codex eating your ChatGPT allowance.",[15,19,20,21,25],{},"OpenAI's help centre confirms the ",[22,23,24],"strong",{},"ChatGPT and Codex shared usage limit",": starting in December, Plus, Pro and Business plans get one allowance covering both ChatGPT and Codex, measured in tokens. Instant chat will not count against it, and API usage stays billed separately. OpenAI has not announced an exact start date yet.",[15,27,28],{},"Picture this. It's a December morning, and you point Codex at a nasty refactor: GPT-6 Astra, Fast mode, three cloud subagents in parallel. The diff is clean.",[15,30,31],{},"Mid-afternoon, you ask ChatGPT to draft a client report. Instead of an answer, you get a reset time.",[15,33,34],{},[35,36,37],"em",{},"Wait. Since when does my coding agent spend my chat budget?",[15,39,40],{},"Since December. Codex already shares an allowance with ChatGPT Work, and now ChatGPT joins that pool. We read every OpenAI page on this (checked 3 November 2026): here's what is confirmed, what isn't, and what we'd do now.",[42,43,45],"h2",{"id":44},"so-what-is-the-chatgpt-and-codex-shared-usage-limit-exactly","So what is the ChatGPT and Codex shared usage limit, exactly?",[15,47,48],{},"OpenAI's help centre article \"Understanding ChatGPT and Codex usage limits\" says it in one sentence: \"Starting in December, Plus, Pro, and Business plans' usage allowance will be shared across ChatGPT and Codex as we bring the agentic capabilities of Work into the main ChatGPT experience.\"",[15,50,51],{},"ChatGPT has three modes: Chat, Work for multi-step jobs, and Codex for software development. According to OpenAI's Codex pricing page, Work and Codex already share one allowance, while Chat has its own message limits.",[15,53,54],{},"So from December, one meter covers almost everything you do.",[15,56,57],{},[58,59],"img",{"alt":60,"src":61},"Before and after diagram of the ChatGPT and Codex shared usage limit in December","/img/blog/chatgpt-codex-shared-usage-limit-before-after.jpg",[63,64,65,81],"table",{},[66,67,68],"thead",{},[69,70,71,75,78],"tr",{},[72,73,74],"th",{},"Plan",[72,76,77],{},"Today",[72,79,80],{},"From December (per OpenAI)",[82,83,84,96,106,116],"tbody",{},[69,85,86,90,93],{},[87,88,89],"td",{},"Plus",[87,91,92],{},"Shared Work and Codex allowance (five-hour and weekly windows). Chat separate.",[87,94,95],{},"One ChatGPT and Codex allowance. Instant chat excluded.",[69,97,98,101,104],{},[87,99,100],{},"Pro ($100, $200, $500)",[87,102,103],{},"Shared Work and Codex allowance, no five-hour limit for now. GPT-6 Pro chat has its own weekly limit.",[87,105,95],{},[69,107,108,111,114],{},[87,109,110],{},"Business",[87,112,113],{},"Shared Work and Codex allowance per seat. Chat separate.",[87,115,95],{},[69,117,118,121,124],{},[87,119,120],{},"Free, Go, Enterprise, Edu",[87,122,123],{},"Codex included with plan-specific limits.",[87,125,126],{},"Not mentioned.",[15,128,129],{},"OpenAI says \"most users won't need to wait for their usage to reset more often than they do now\". Maybe so. But \"most users\" is doing a lot of work in that sentence.",[131,132,133],"blockquote",{},[15,134,135],{},"If you run Codex hard, you are not \"most users\". Plan as if the shared pool will run dry, and enjoy the surprise if it doesn't.",[15,137,138],{},"Still unpublished: the exact day, the new allowance size on each plan, whether reset windows change, and whether the GPT-6 Pro weekly chat limit stays separate. Treat any date or number you see elsewhere as a guess until OpenAI publishes it.",[42,140,142],{"id":141},"which-plans-get-pulled-into-one-pool","Which plans get pulled into one pool",[15,144,145],{},"Current prices, per OpenAI's Codex pricing page:",[147,148,149,153,156],"ul",{},[150,151,152],"li",{},"Plus: $20 a month.",[150,154,155],{},"Pro: $100, $200 or $500 a month, with Ultrafast mode only on Pro $500.",[150,157,158],{},"Business: $20 per user per month billed annually (two or more users), or $25 billed monthly.",[15,160,161,162,165],{},"Here's the weird part. ",[22,163,164],{},"Pro users got a separate, friendlier change first."," According to OpenAI's ChatGPT release notes, on 29 October Pro $100 and Pro $500 started including twice their previous Work and Codex allowance.",[15,167,168],{},"OpenAI also cancelled a Pro $200 reduction announced on 29 September and doubled the lower Pro $200 allowance, so every Pro $200 subscriber has the higher one, with no wait for renewal.",[15,170,171],{},"Two caveats from OpenAI's Pro help article: Plus usage is unchanged, and \"a larger allowance does not guarantee twice as many completed tasks\".",[15,173,174],{},"Our read: on Plus, nothing got bigger, but more things will soon drink from the same glass.",[42,176,178],{"id":177},"what-counts-against-the-shared-pool-and-what-costs-8x","What counts against the shared pool (and what costs 8x)",[15,180,181],{},"OpenAI hasn't published a full list for December. The only named exclusion is Instant chat.",[15,183,184],{},"Here's what draws from the Work and Codex allowance today, per OpenAI's Codex pricing page and docs:",[63,186,187,197],{},[66,188,189],{},[69,190,191,194],{},[72,192,193],{},"Activity",[72,195,196],{},"How it counts today",[82,198,199,207,215,223,231,239,247,255],{},[69,200,201,204],{},[87,202,203],{},"Local Codex messages (CLI, IDE, app)",[87,205,206],{},"Draws from the shared allowance",[69,208,209,212],{},[87,210,211],{},"Cloud tasks",[87,213,214],{},"\"May use more of your allowance than local messages\"",[69,216,217,220],{},[87,218,219],{},"Subagents",[87,221,222],{},"\"Consume more tokens than comparable single-agent runs\"",[69,224,225,228],{},[87,226,227],{},"Fast mode",[87,229,230],{},"2.5x the Standard rate",[69,232,233,236],{},[87,234,235],{},"Ultrafast (Pro $500)",[87,237,238],{},"8x on GPT-6 Astra and GPT-6.1 Sol",[69,240,241,244],{},[87,242,243],{},"Image generation",[87,245,246],{},"Uses your allowance about 3 to 5x faster",[69,248,249,252],{},[87,250,251],{},"Auto-review safety checks",[87,253,254],{},"Extra model calls that can add to usage",[69,256,257,260],{},[87,258,259],{},"Instant chat (from December)",[87,261,262],{},"Will not count",[15,264,265],{},[58,266],{"alt":267,"src":268},"What counts against the ChatGPT and Codex shared usage limit: speed multipliers and five-hour estimates by model","/img/blog/chatgpt-codex-shared-usage-limit-burn-rate.jpg",[15,270,271,274],{},[22,272,273],{},"The model matters more than anything in that table."," Per five hours on Plus, OpenAI estimates 5 to 45 local messages on GPT-6 Astra, 15 to 160 on GPT-6.1 Sol, 35 to 285 on GPT-6.1 Terra and 350 to 3,000 on GPT-6 Luna.",[15,276,277],{},"Those are estimates, not fixed limits, and weekly limits may also apply. Still, look at the spread: Luna's low estimate is higher than Astra's high one.",[131,279,280],{},[15,281,282],{},"Astra in Fast mode with three cloud subagents is how you see \"codex limit reached\" before lunch. Pick the model before you write the prompt.",[42,284,286],{"id":285},"how-to-check-your-codex-and-chatgpt-usage-before-it-bites","How to check your Codex and ChatGPT usage before it bites",[15,288,289],{},"Three places show it:",[291,292,293,296,299],"ol",{},[150,294,295],{},"ChatGPT, Settings → Usage: remaining allowance and reset times, combined from December.",[150,297,298],{},"Codex app, Usage & Billing: the same allowance plus any credit balance.",[150,300,301,302,306],{},"Codex CLI: run ",[303,304,305],"code",{},"/status"," inside a session.",[308,309,314],"pre",{"className":310,"code":311,"language":312,"meta":313,"style":313},"language-bash shiki shiki-themes github-light","# inside a running Codex CLI session\n/status\n","bash","",[303,315,316,325],{"__ignoreMap":313},[317,318,321],"span",{"class":319,"line":320},"line",1,[317,322,324],{"class":323},"sAwPA","# inside a running Codex CLI session\n",[317,326,328],{"class":319,"line":327},2,[317,329,331],{"class":330},"s7eDp","/status\n",[15,333,334,335,338],{},"Now the reset logic, which trips people up. On Plus and Standard Business there's a five-hour window and a weekly window, and ",[22,336,337],{},"you need allowance left in both",".",[15,340,341],{},"According to OpenAI's help centre, a new five-hour window starts with your first message after the previous one ends, not on a fixed clock. So you can hit the five-hour cap while weekly allowance remains. Pro currently has no five-hour limit.",[15,343,344],{},[58,345],{"alt":346,"src":347},"How to check Codex usage and how the five-hour and weekly windows of the shared usage limit reset","/img/blog/chatgpt-codex-shared-usage-limit-check-usage.jpg",[15,349,350],{},"OpenAI suggests checking every week or two. We'd check daily in the first week of December. Boring advice, and the cheapest insurance you'll ever buy.",[42,352,354],{"id":353},"_6-ways-to-make-the-chatgpt-and-codex-shared-usage-limit-last","6 ways to make the ChatGPT and Codex shared usage limit last",[15,356,357],{},"Every tip below comes from OpenAI's own usage tips or credit rates.",[15,359,360],{},[58,361],{"alt":362,"src":363},"Six ways to make the ChatGPT and Codex shared usage limit last, with credit rates","/img/blog/chatgpt-codex-shared-usage-limit-six-ways.jpg",[15,365,366],{},[22,367,368],{},"1. Keep AGENTS.md short",[15,370,371],{},"Codex reads AGENTS.md before doing any work, so every line costs tokens on every run. OpenAI's tips literally say \"Reduce the size of your AGENTS.md\", and Codex stops loading instruction files at 32 KiB combined by default.",[15,373,374,375,380],{},"Keep the root file to stack, commands and hard rules. Our ",[376,377,379],"a",{"href":378},"/blog/agents-md-best-practices","AGENTS.md best practices"," guide has a short template.",[382,383,385],"h3",{"id":384},"_2-put-sub-tasks-on-a-smaller-model","2. Put sub-tasks on a smaller model",[15,387,388],{},"In OpenAI's credit table, GPT-6 Luna costs 2.5 credits per million input tokens. GPT-6.1 Sol costs 50, and GPT-6 Astra costs 250. OpenAI's subagent docs suggest Luna for \"fast, narrowly scoped\" work, so make it the default for spawned agents:",[308,390,395],{"className":391,"code":393,"language":394,"meta":313},[392],"language-text","# .codex/config.toml\n[agents]\ndefault_subagent_model = \"gpt-6-luna\"\nmax_concurrent_threads_per_session = 2\n","text",[303,396,393],{"__ignoreMap":313},[15,398,399],{},"Reasoning effort matters too. OpenAI says higher effort \"can improve results for complex tasks, but it takes longer and uses more tokens\", so start low and raise it only when a task needs it.",[382,401,403],{"id":402},"_3-run-fewer-parallel-background-agents","3. Run fewer parallel background agents",[15,405,406,407,410],{},"Each subagent does its own model and tool work, so parallel runs multiply usage. The ",[303,408,409],{},"max_concurrent_threads_per_session"," setting above caps them.",[15,412,413],{},"Go parallel when the work splits cleanly, not because five agents look cool in the terminal.",[382,415,417],{"id":416},"_4-keep-context-cached","4. Keep context cached",[15,419,420],{},"OpenAI charges 25 credits per million cached input tokens on GPT-6 Astra instead of 250, and 2.5 instead of 50 on GPT-6.1 Sol. Same tokens, a fraction of the price.",[15,422,423],{},"Caching needs an identical prompt start, so continue threads instead of re-pasting files, and don't edit AGENTS.md mid-session.",[382,425,427],{"id":426},"_5-scope-tasks-tightly","5. Scope tasks tightly",[15,429,430],{},"OpenAI's tips: precise prompts, only the relevant files, a defined output, and no unused MCP servers, because \"every MCP server adds more context to your messages\". Skip Fast mode unless speed matters.",[15,432,433,434,338],{},"For more ways to cut tokens per call, see our list of ",[376,435,437],{"href":436},"/blog/skills-that-reduce-token-usage","skills that reduce token usage",[382,439,441],{"id":440},"_6-move-heavy-automation-to-an-api-key","6. Move heavy automation to an API key",[15,443,444],{},"OpenAI calls Codex with an API key \"great for automation in shared environments like CI\". Those sessions are billed at API pricing and leave your plan allowance alone.",[15,446,447,448,451],{},"Move CI jobs and scripted ",[303,449,450],{},"codex exec"," runs to a key. The catch: no cloud features such as GitHub code review or Slack.",[42,453,455],{"id":454},"when-an-api-key-beats-a-chatgpt-plan","When an API key beats a ChatGPT plan",[15,457,458],{},"A plan is a flat fee for an allowance that resets. An API key has no five-hour or weekly window: you pay per token, within the provider's rate limits.",[15,460,461],{},"This is where most people get it wrong. They compare a $20 plan to a per-token bill, when the real question is what kind of work you do.",[15,463,464],{},"A key tends to win for automated, steady work (CI, scheduled jobs) that can run on a small model. A plan tends to win for interactive use on premium models like GPT-6 Astra, and cloud features need one anyway.",[15,466,467],{},"One caution from OpenAI: API token prices \"are separate from subscription usage; don't use them to estimate included tasks\". So run your real workload on a key for a week and compare the bill.",[15,469,470,471,475],{},"Since 28 October, OpenAI offers promotional API credits to eligible Pro and Business subscribers, which can cover that test. Our comparison of ",[376,472,474],{"href":473},"/blog/openrouter-vs-direct-api-agents","OpenRouter vs direct API keys for agents"," covers whether to call OpenAI directly or through a router.",[15,477,478,479],{},"That split, plans for humans and keys for robots, is why we built BetterClaw the way we did. Run your agent on your own API keys with zero markup across 28+ providers, so a plan limit never stops it. ",[376,480,484],{"href":481,"rel":482},"https://app.betterclaw.io/sign-in",[483],"nofollow","Start your agent in 60 seconds",[42,486,488],{"id":487},"codex-vs-claude-code-vs-your-own-agent-whose-limits-bite","Codex vs Claude Code vs your own agent: whose limits bite?",[15,490,491],{},"Switching tools helps less than you'd think. According to Anthropic's help centre, Claude Pro and Max plans have a five-hour session limit and a weekly limit, shared across Claude and Claude Code.",[15,493,494],{},"Same shape as Codex, different logo on the meter.",[63,496,497,512],{},[66,498,499],{},[69,500,501,503,506,509],{},[72,502],{},[72,504,505],{},"Codex on a ChatGPT plan",[72,507,508],{},"Claude Code on a Claude plan",[72,510,511],{},"Your own agent on API keys",[82,513,514,528,542],{},[69,515,516,519,522,525],{},[87,517,518],{},"Limit type",[87,520,521],{},"Plan allowance, shared with ChatGPT from December",[87,523,524],{},"Plan limits shared across Claude and Claude Code",[87,526,527],{},"Provider rate limits and your spend",[69,529,530,533,536,539],{},[87,531,532],{},"Reset window",[87,534,535],{},"Five-hour and weekly on Plus and Business; no five-hour on Pro for now",[87,537,538],{},"Five-hour and weekly on Pro and Max; weekly only on verified Ultra",[87,540,541],{},"No plan window",[69,543,544,547,550,553],{},[87,545,546],{},"At the cap",[87,548,549],{},"Current turn can finish, then wait, upgrade, buy credits or use an API key",[87,551,552],{},"Warnings, step may wrap up, then wait or use usage credits",[87,554,555],{},"Rate-limit errors, or a stop at your spend limit",[15,557,558],{},[58,559],{"alt":560,"src":561},"Codex vs Claude Code vs your own agent: usage limits and reset windows compared","/img/blog/chatgpt-codex-shared-usage-limit-limits-compared.jpg",[15,563,564],{},"Both big coding tools now tie your coding agent to your chat allowance. Moving to Claude Code gives you a second allowance, not a different kind of one.",[15,566,567],{},"An agent on your own keys is the only option here with no plan window. It isn't automatically cheaper, but a long Codex session can never lock it out.",[15,569,570,571,575],{},"If limits are why you're shopping around, read our roundup of ",[376,572,574],{"href":573},"/blog/claude-code-chatgpt-rate-limit-alternatives-2026","Claude Code and ChatGPT rate limit alternatives"," before you move anything.",[42,577,579],{"id":578},"what-december-is-really-telling-you","What December is really telling you",[15,581,582],{},"A subscription allowance is rented capacity. OpenAI decides its size, its windows and, from December, what else shares it.",[15,584,585],{},"That's not a scandal. It's how plans work, and for interactive coding a flat fee is still hard to beat.",[15,587,588],{},"What changes is the blast radius. One greedy afternoon in Codex now shows up in the chat you open at dinner.",[131,590,591],{},[15,592,593],{},"Give interactive work the plan. Give unattended work its own key. Do that in November and December barely registers.",[595,596,598,603,606],"callout",{"type":597},"cta",[15,599,600],{},[22,601,602],{},"Keep your agent off the shared meter",[15,604,605],{},"BetterClaw runs your agent on your own API keys, zero markup across 28+ providers, with 500+ connectors, so a ChatGPT reset never decides when your automation runs. Pro is $49 ($39 billed annually), Business $149, Enterprise custom.",[15,607,608,609],{},"Plans start at $19 a month on Starter, with a 7-day free trial on your first workspace. ",[376,610,612],{"href":611},"/pricing","See plans and pricing",[42,614,616],{"id":615},"frequently-asked-questions","Frequently Asked Questions",[382,618,620],{"id":619},"what-is-the-chatgpt-and-codex-shared-usage-limit","What is the ChatGPT and Codex shared usage limit?",[15,622,623],{},"It's one allowance covering both ChatGPT and Codex on Plus, Pro and Business plans from December, measured in tokens. Instant chat will not draw from it.",[382,625,627],{"id":626},"when-does-the-shared-limit-start-and-which-plans-does-it-affect","When does the shared limit start, and which plans does it affect?",[15,629,630],{},"OpenAI says \"starting in December\" with no exact date as of 3 November 2026. Its notice names only Plus, Pro and Business.",[382,632,634],{"id":633},"how-do-i-check-how-much-codex-has-used","How do I check how much Codex has used?",[15,636,637,638,640],{},"Open Settings → Usage in ChatGPT or Usage & Billing in the Codex app. In the Codex CLI, run ",[303,639,305],{}," during a session.",[382,642,644],{"id":643},"will-hitting-the-limit-break-a-codex-task-halfway-through","Will hitting the limit break a Codex task halfway through?",[15,646,647],{},"No. Codex can finish the current turn, subject to fair use. Then you wait for the reset, upgrade, buy credits, or run local Codex chats on an API key.",[382,649,651],{"id":650},"is-an-api-key-cheaper-than-a-chatgpt-plan-for-codex","Is an API key cheaper than a ChatGPT plan for Codex?",[15,653,654],{},"It depends on the work: keys tend to win for steady automation on small models, plans for interactive use on premium models. OpenAI says not to use API prices to estimate included tasks, so test your real workload for a week.",[382,656,658],{"id":657},"is-claude-code-a-good-alternative-to-codex-for-limits","Is Claude Code a good alternative to Codex for limits?",[15,660,661,662,338],{},"As a second allowance, yes. As an escape from limits, no: Anthropic's Pro and Max plans also have a five-hour session limit and a weekly limit, shared between Claude and Claude Code. If you're already stuck, see our fix guide for ",[376,663,665],{"href":664},"/blog/claude-rate-limit-reached","Claude rate limit reached",[667,668,669],"style",{},"html pre.shiki code .sAwPA, html code.shiki .sAwPA{--shiki-default:#6A737D}html pre.shiki code .s7eDp, html code.shiki .s7eDp{--shiki-default:#6F42C1}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}",{"title":313,"searchDepth":327,"depth":327,"links":671},[672,673,674,675,676,684,685,686,687],{"id":44,"depth":327,"text":45},{"id":141,"depth":327,"text":142},{"id":177,"depth":327,"text":178},{"id":285,"depth":327,"text":286},{"id":353,"depth":327,"text":354,"children":677},[678,680,681,682,683],{"id":384,"depth":679,"text":385},3,{"id":402,"depth":679,"text":403},{"id":416,"depth":679,"text":417},{"id":426,"depth":679,"text":427},{"id":440,"depth":679,"text":441},{"id":454,"depth":327,"text":455},{"id":487,"depth":327,"text":488},{"id":578,"depth":327,"text":579},{"id":615,"depth":327,"text":616,"children":688},[689,690,691,692,693,694],{"id":619,"depth":679,"text":620},{"id":626,"depth":679,"text":627},{"id":633,"depth":679,"text":634},{"id":643,"depth":679,"text":644},{"id":650,"depth":679,"text":651},{"id":657,"depth":679,"text":658},"Guides","2026-11-06","ChatGPT and Codex shared usage limit starts in December. See which plans change, what counts against the pool, and 6 ways to make your allowance last.","md",false,"/img/blog/chatgpt-codex-shared-usage-limit.jpg",null,{},true,"/blog/chatgpt-codex-shared-usage-limit","11 min read",{"title":5,"description":697},"ChatGPT and Codex Shared Usage Limit: What Changes and When","blog/chatgpt-codex-shared-usage-limit",[710,711,712,713,714,715,716],"chatgpt codex shared usage limit","codex usage limit","chatgpt codex limit","codex limit reached","chatgpt plus codex limit","codex pro limits","codex rate limit reset","KmphP870u_tJ6DqmfKPW2vnRKxJMf2J1tDvRpEwjsQM",[719,726,733],{"stem":720,"title":721,"description":722,"date":696,"image":723,"category":724,"redirected":699,"author":725,"readingTime":705},"blog/glm-5-3-vs-mimo-v2-6-pro","GLM 5.3 vs MiMo v2.6 Pro: Which Model Should Run Your Agent?","GLM 5.3 vs MiMo v2.6 Pro compared for AI agents: pricing, coding and tool-calling benchmarks, open weights and which model to pick for your workload.","/img/blog/glm-5-3-vs-mimo-v2-6-pro.jpg","Comparison",{"name":7,"role":8,"avatar":9,"bio":10},{"stem":727,"title":728,"description":729,"date":696,"image":730,"category":695,"redirected":699,"author":731,"readingTime":732},"blog/lfm3-ollama-tool-calling-agent-setup","LFM3 on Ollama: Setup, Tool Calling and Agent Fixes","Run LFM3 on Ollama from Liquid AI: which tag to pull, how much memory it needs, whether tool calling works, and the fix when tool calls come back as text.","/img/blog/lfm3-ollama-tool-calling-agent-setup.jpg",{"name":7,"role":8,"avatar":9,"bio":10},"13 min read",{"stem":734,"title":735,"description":736,"date":737,"image":738,"category":695,"redirected":699,"author":739,"readingTime":732},"blog/deepseek-v4-1-flash-agents","DeepSeek V4.1 Flash 1017 for AI Agents: Price, Hardware and Local vs Ollama Cloud vs API","DeepSeek V4.1 Flash 1017 costs $0.30/$1.20 per 1M tokens at peak, half off-peak. Local vs Ollama Cloud vs API, hardware needs and agent setup.","2026-11-05","/img/blog/deepseek-v4-1-flash-agents.jpg",{"name":7,"role":8,"avatar":9,"bio":10},1793972859874]