{"id":6974,"date":"2026-11-07T07:24:26","date_gmt":"2026-11-07T07:24:26","guid":{"rendered":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents"},"modified":"2026-11-07T07:24:26","modified_gmt":"2026-11-07T07:24:26","slug":"openai-biobasicsbench-wet-lab-ai-agents","status":"publish","type":"post","link":"https:\/\/i10x.ai\/news\/openai-biobasicsbench-wet-lab-ai-agents","title":{"rendered":"OpenAI&#8217;s BioBasicsBench Tests AI Agents on Real Wet-Lab Data"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">OpenAI wants to know whether AI agents can do the unglamorous part of biology: reading a smudged gel, noticing a mislabeled control plate, or admitting that the data cannot support a conclusion. On November 6, 2026, the company <a href=\"https:\/\/openai.com\/index\/measuring-how-ai-agents-handle-wet-lab-science\" target=\"_blank\" rel=\"noreferrer noopener\">introduced BioBasicsBench<\/a>, a benchmark of 100 routine wet-lab tasks developed with Ligo, a biotech company focused on enzyme design. Its best model, GPT-6 Astra, earned 66.7% of the available credit at the highest common reasoning setting. OpenAI&#8217;s own write-up is just as interested in where Astra still fails.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u26a1 Quick Take<\/h2>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>What:<\/strong> BioBasicsBench tests whether agents can interpret raw experimental data, catch errors and recognize insufficient evidence across 100 tasks from real Ligo experiments.<\/li><li><strong>Results:<\/strong> At xhigh effort, GPT-6 Astra scored 66.7% mean reward, GPT-6.1 Sol 62.0% and GPT-5.6 Sol 43.8% \u2014 with Astra using about half the output tokens of GPT-5.6 Sol.<\/li><li><strong>Caveats:<\/strong> Scores are partial rubric credit, not task success rates; only OpenAI models were tested; and the tasks have not been released.<\/li><li><strong>Why it matters:<\/strong> Lab AI has to get everyday evidence right before it can speed up discovery \u2014 and OpenAI says the benchmark will also inform biosafety and biosecurity tracking.<\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">What happened<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">OpenAI announced the benchmark on its blog and <a href=\"https:\/\/x.com\/OpenAI\/status\/2118782708767965238\" target=\"_blank\" rel=\"noreferrer noopener\">on X<\/a>, writing that &#8220;Accelerating biological research starts with getting the everyday work right.&#8221; Ligo CEO <a href=\"https:\/\/x.com\/edharris\/status\/2118790408398606784\" target=\"_blank\" rel=\"noreferrer noopener\">Ed Harris<\/a> said the benchmark &#8220;reveals where models still struggle with routine experimental tasks.&#8221; According to OpenAI, existing benchmarks built around &#8220;charts and images selected for scientific papers may miss&#8221; the ambiguity of real bench work \u2014 failed experiments, ambiguous images and conflicting observations. BioBasicsBench is meant to measure exactly that.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Background: OpenAI&#8217;s push into science<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The release fits a broader science push at OpenAI. <a href=\"https:\/\/www.rdworldonline.com\/openai-unveils-biobasicsbench-a-benchmark-to-gauge-how-well-ai-agents-read-raw-experimental-data-across-100-routine-wet-lab-tasks\/\" target=\"_blank\" rel=\"noreferrer noopener\">R&#038;D World<\/a> notes that CEO Sam Altman said in a November 3 podcast that the company&#8217;s recent math results &#8220;really made me think that we are closer to this milestone than I expected us to be,&#8221; referring to a machine for scientific discovery. Days earlier, OpenAI president Greg Brockman <a href=\"https:\/\/i10x.ai\/news\/openai-brockman-ai-disease-cures-12-18-months\">forecast significant AI strides in biology and disease cures within 12 to 18 months<\/a>. GPT-6 Astra, OpenAI&#8217;s top tier released in early September according to R&#038;D World, has been at the center of several recent claims, including <a href=\"https:\/\/i10x.ai\/news\/weight-space-programming-gpt-6-astra-qwen-edit\">an MIT result on editing other models&#8217; weights<\/a>. BioBasicsBench is a more sober exercise: it measures routine competence rather than breakthroughs.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">How BioBasicsBench works<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The 100 tasks span five areas of routine wet-lab work: protein methods, imaging and cytometry, quantification and assays, nucleic acid methods, and molecular cloning. They range from reading the molecular weight of a DNA sample on a gel to detailed assay analysis. Ligo scientists wrote each task from their own experimental data and paired it with an expert-written rubric. Every task went through an internal review by a Ligo scientist who had not written it, an external review by a PhD-level wet-lab scientist, and a final internal review.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Inputs preserve the messiness of real lab evidence. By OpenAI&#8217;s count, 31 tasks include gels and blots, 27 protocols and notes, 20 tables and instrument exports, 18 bench and reagent photos, 14 microscopy data, 13 instrument screens and plots, 12 sequencing files, 7 protocol videos and 7 agar plates. Agents return a structured answer \u2014 a conclusion, measurement or recommendation, plus reasoning where required \u2014 and earn partial credit per rubric criterion, including for recognizing that &#8220;the evidence is insufficient.&#8221;<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">OpenAI evaluated seven of its own models across 37 model-and-effort settings, five attempts per task, for 18,500 results. The headline metric is mean reward: the average share of available credit, with 95% confidence intervals from resampling whole tasks.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Model (xhigh effort)<\/th><th>Mean reward<\/th><th>95% CI<\/th><th>Output tokens \/ task<\/th><\/tr><\/thead><tbody><tr><td>GPT-6 Astra<\/td><td>66.7%<\/td><td>60.3\u201373.1%<\/td><td>4,704<\/td><\/tr><tr><td>GPT-6.1 Sol<\/td><td>62.0%<\/td><td>55.6\u201368.4%<\/td><td>5,238<\/td><\/tr><tr><td>GPT-5.6 Sol<\/td><td>43.8%<\/td><td>37.6\u201350.0%<\/td><td>9,099<\/td><\/tr><tr><td>GPT-6 Luna<\/td><td>38.3%<\/td><td>32.3\u201344.5%<\/td><td>13,120<\/td><\/tr><tr><td>GPT-5.5<\/td><td>35.4%<\/td><td>29.7\u201341.0%<\/td><td>11,135<\/td><\/tr><tr><td>GPT-5.4<\/td><td>33.2%<\/td><td>28.3\u201338.4%<\/td><td>13,717<\/td><\/tr><tr><td>GPT-5.2<\/td><td>31.2%<\/td><td>26.2\u201336.3%<\/td><td>27,620<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\">What Astra gets right \u2014 and wrong<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">OpenAI credits GPT-6 Astra&#8217;s gains to &#8220;stronger native multimodal capabilities.&#8221; In one cloning task, GPT-5.6 Sol accepted an empty negative-control plate at face value; Astra noticed the rim label read &#8220;LB Kan50&#8221; \u2014 kanamycin, not the ampicillin used in the experiment \u2014 and rejected the control. In another, Astra recognized a microscope z-stack where GPT-5.6 Sol searched for a biological reason why cells seemed to vanish. Astra also refused to call cells dead from a dye signal when overlapping detection channels made brightness alone inconclusive.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The failure examples are equally instructive. In a colony PCR task, Astra grouped the gel lanes correctly but &#8220;assumes that a larger PCR product would identify successful cloning,&#8221; picking colonies 4, 6, 7 and 8 instead of the reference 5, 9 and 10 \u2014 which would lead to purifying the wrong plasmid. It generalized that vacuoles look darker while ignoring lighter ones in the same image. And it overcounted one colony plate by about 12%, which would inflate an estimate of transformation efficiency by the same proportion. OpenAI&#8217;s conclusion: &#8220;Higher benchmark scores do not yet mean models can make reliable judgments in the lab.&#8221;<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Pros and cons<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">For users (scientists)<\/h3>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Pro:<\/strong> A benchmark built on raw gels, plates and videos is far closer to daily lab work than paper-figure quizzes, and it rewards saying &#8220;not enough evidence.&#8221;<\/li><li><strong>Con:<\/strong> A two-thirds credit score still means frequent partial errors on routine tasks; scientists cannot yet hand off interpretation without checking.<\/li><\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">For developers<\/h3>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Pro:<\/strong> The results suggest multimodal perception, not longer reasoning, drives gains \u2014 Astra scored higher with fewer tokens than GPT-5.6 Sol.<\/li><li><strong>Con:<\/strong> The tasks are not public, so outside teams cannot reproduce results, test other models or build against the benchmark directly.<\/li><\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">For companies (biotech and pharma)<\/h3>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>Pro:<\/strong> Rubric-based scores give lab leaders a measurable acceptance criterion before letting agents into analysis pipelines.<\/li><li><strong>Con:<\/strong> With only OpenAI models tested, procurement teams get no head-to-head comparison with Anthropic, Google or open-weight alternatives.<\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">What X is saying<\/h2>\n\n\n\n<blockquote class=\"twitter-tweet i10x-x-embed\"><p lang=\"en\" dir=\"ltr\">A new way to measure how AI agents handle the routine analyses of wet-lab science\u2014BioBasicsBench. Developed with Ligo, a biotech company focused on enzyme design, the benchmark draws on 100 tasks from real experiments to test whether agents can interpret raw lab data, identify errors, and recognize when the evidence doesn\u2019t support a conclusion.<\/p>&mdash; OpenAI (@OpenAI) <a href=\"https:\/\/x.com\/OpenAI\/status\/2118782708767965238\" target=\"_blank\" rel=\"noreferrer noopener\">November 6, 2026 \u00b7 View on X<\/a><\/blockquote>\n\n\n\n<blockquote class=\"twitter-tweet i10x-x-embed\"><p lang=\"en\" dir=\"ltr\">Excited to release BioBasicsBench, a benchmark we developed with @OpenAI to test whether AI agents can interpret real lab data, spot experimental errors, and recognize when the evidence doesn\u2019t support a conclusion.<\/p>&mdash; Ed Harris (@edharris) <a href=\"https:\/\/x.com\/edharris\/status\/2118790408398606784\" target=\"_blank\" rel=\"noreferrer noopener\">November 6, 2026 \u00b7 View on X<\/a><\/blockquote>\n\n\n\n<blockquote class=\"twitter-tweet i10x-x-embed\"><p lang=\"en\" dir=\"ltr\">At the xhigh setting, Astra used 4,704 output tokens per task, about half the older model\u2019s, so longer reasoning didn\u2019t drive the gain.<\/p>&mdash; Rohan Paul (@rohanpaul_ai) <a href=\"https:\/\/x.com\/rohanpaul_ai\/status\/2118794080885051683\" target=\"_blank\" rel=\"noreferrer noopener\">November 6, 2026 \u00b7 View on X<\/a><\/blockquote>\n\n\n\n<p class=\"wp-block-paragraph\">Several observers zoomed in on the scoring. &#8220;OpenAI&#8217;s top reported score is 66.7% with partial credit\u2014not a task success rate,&#8221; <a href=\"https:\/\/x.com\/haamzaC\/status\/2118810736101126606\" target=\"_blank\" rel=\"noreferrer noopener\">Microsoft cloud and AI director Hamza Chishti<\/a> wrote. <a href=\"https:\/\/x.com\/Om_Dream1\/status\/2118793279114244126\" target=\"_blank\" rel=\"noreferrer noopener\">Software engineer Omar<\/a> liked that the benchmark tests for &#8220;there isn&#8217;t enough evidence&#8221; and added: &#8220;I&#8217;d love to see that ability broken out from the overall score.&#8221; The launch also spawned a viral hoax: a <a href=\"https:\/\/x.com\/dummerspast39\/status\/2118823037331968510\" target=\"_blank\" rel=\"noreferrer noopener\">post<\/a> with more than 750 likes claimed a browser-console command revealed a hidden &#8220;GPT-6.2 Astra&#8221; line at 89%. OpenAI&#8217;s post lists seven tested models, and no such model appears in its results. Analyst <a href=\"https:\/\/x.com\/MichaelWaitze\/status\/2118853283313857022\" target=\"_blank\" rel=\"noreferrer noopener\">Michael Waitze<\/a> noted the expectation gap more soberly: &#8220;We actually went into why today&#8217;s release was BioBasicsBench, not Astra.&#8221; For background on the next Astra version, see <a href=\"https:\/\/i10x.ai\/news\/gpt-6-2-astra-sol-google-cloud-leak\">our report on GPT-6.2 Astra and Sol sightings<\/a>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What Grok says<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Asked about the hidden-model claim, <a href=\"https:\/\/x.com\/grok\/status\/2118841531708702920\" target=\"_blank\" rel=\"noreferrer noopener\">Grok<\/a> called it &#8220;engagement bait,&#8221; describing a recipe in which a video injects &#8220;a fake window.__bioBasicsBench.reveal call&#8221; so that &#8220;a nonexistent GPT-6.2 Astra line at 89% pops up.&#8221; That is consistent with OpenAI&#8217;s published results, which include only seven models.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Critical analysis<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">BioBasicsBench gets the target right. Most science benchmarks test knowledge or paper comprehension; this one tests whether a model can be trusted with the boring evidence that decides whether an experiment worked. Rewarding &#8220;the data cannot support this&#8221; is especially valuable, because over-interpretation is a classic failure for both humans and models.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The limits are just as clear. OpenAI built, ran and reported the benchmark and evaluated only its own models; BenchLM notes OpenAI &#8220;has not released the tasks,&#8221; and keeps the scores out of its weighted rankings. One hundred tasks from one company&#8217;s lab is a narrow slice of biology, and rubric credit blurs the difference between a minor slip and a decision that wastes a week of bench time \u2014 such as selecting the wrong colonies. External replication and competitor results will determine whether this becomes a field standard or stays a vendor showcase. OpenAI also frames the benchmark as a way to track biosafety and biosecurity implications; it has not said how scores feed into its safeguards.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What&#8217;s still open<\/h2>\n\n\n\n<ul class=\"wp-block-list\"><li>Will OpenAI and Ligo release the tasks or an evaluation harness for outside models?<\/li><li>How do Claude, Gemini and open-weight models score on the same tasks?<\/li><li>How does performance break down for the &#8220;insufficient evidence&#8221; criterion specifically?<\/li><li>How will BioBasicsBench results inform OpenAI&#8217;s biosafety thresholds?<\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">\u270d\ufe0f About the analysis<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">This independent, research-based i10x analysis is based on OpenAI&#8217;s November 6, 2026 BioBasicsBench announcement, R&#038;D World&#8217;s coverage, BenchLM&#8217;s benchmark page, and posts on X from OpenAI, Ligo and independent observers. All scores are as reported by OpenAI and have not been independently reproduced. More on <a href=\"https:\/\/i10x.ai\/news\/company\/open-ai\">OpenAI<\/a>.<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><a href=\"https:\/\/openai.com\/index\/measuring-how-ai-agents-handle-wet-lab-science\" target=\"_blank\" rel=\"noreferrer noopener\">OpenAI: Measuring how AI agents handle everyday wet-lab science<\/a><\/li><li><a href=\"https:\/\/www.rdworldonline.com\/openai-unveils-biobasicsbench-a-benchmark-to-gauge-how-well-ai-agents-read-raw-experimental-data-across-100-routine-wet-lab-tasks\/\" target=\"_blank\" rel=\"noreferrer noopener\">R&#038;D World: OpenAI unveils BioBasicsBench<\/a><\/li><li><a href=\"https:\/\/benchlm.ai\/benchmarks\/biobasicsbench\" target=\"_blank\" rel=\"noreferrer noopener\">BenchLM: BioBasicsBench<\/a><\/li><li><a href=\"https:\/\/x.com\/OpenAI\/status\/2118782708767965238\" target=\"_blank\" rel=\"noreferrer noopener\">OpenAI announcement on X<\/a><\/li><li><a href=\"https:\/\/x.com\/edharris\/status\/2118790408398606784\" target=\"_blank\" rel=\"noreferrer noopener\">Ligo CEO Ed Harris on X<\/a><\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83d\udd2d i10x Perspective<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The road to AI-accelerated biology runs through the lab bench, not the press release. BioBasicsBench is a useful honesty check: even OpenAI&#8217;s best model misreads controls and miscounts colonies. The next milestone is not a higher score from one vendor, but open tasks, rival models on the same leaderboard, and lab teams measuring agents against their own data before trusting them.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>OpenAI and enzyme-design biotech Ligo released BioBasicsBench, a benchmark of 100 routine wet-lab tasks built from raw gels, colony plates, microscopy, sequencing files and bench videos. At xhigh reasoning effort, GPT-6 Astra earned 66.7% of available credit, ahead of GPT-6.1 Sol at 62.0% and GPT-5.6 Sol at 43.8%. OpenAI&#8217;s own examples show where Astra still gets lab basics wrong \u2014 and the tasks themselves are not public.<\/p>\n","protected":false},"author":2,"featured_media":6973,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"paper_title":"","paper_url":"","pdf_url":"","doi":"","arxiv_id":"","authors":"","institutions":"","paper_date":"","venue":"","peer_reviewed":false,"code_url":"","data_url":"","funding_coi":"","figure_license":"","paper_license":"","replication":"","study_models":"","study_sample":"","footnotes":"","_yoast_wpseo_primary_category":"2"},"categories":[14,9,2],"tags":[],"companies":[38],"products":[],"class_list":["post-6974","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-agents","category-models","category-open-ai","company-open-ai"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.7 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>BioBasicsBench: OpenAI Tests AI Agents on Wet-Lab Data<\/title>\n<meta name=\"description\" content=\"BioBasicsBench, built by OpenAI and Ligo, tests AI agents on 100 routine wet-lab tasks. GPT-6 Astra earns 66.7% of credit and still miscounts colonies.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"BioBasicsBench: OpenAI Tests AI Agents on Wet-Lab Data\" \/>\n<meta property=\"og:description\" content=\"BioBasicsBench, built by OpenAI and Ligo, tests AI agents on 100 routine wet-lab tasks. GPT-6 Astra earns 66.7% of credit and still miscounts colonies.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents\" \/>\n<meta property=\"og:site_name\" content=\"i10X News\" \/>\n<meta property=\"article:published_time\" content=\"2026-11-07T07:24:26+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/11\/openai-biobasicsbench-title.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"1344\" \/>\n\t<meta property=\"og:image:height\" content=\"768\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Christopher Ort\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@i10x_ai\" \/>\n<meta name=\"twitter:site\" content=\"@i10x_ai\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Christopher Ort\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents\"},\"author\":{\"name\":\"Christopher Ort\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/person\\\/d410bb7ee72febefca2c9c369939219a\"},\"headline\":\"OpenAI&#8217;s BioBasicsBench Tests AI Agents on Real Wet-Lab Data\",\"datePublished\":\"2026-11-07T07:24:26+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents\"},\"wordCount\":1619,\"publisher\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/openai-biobasicsbench-title.webp\",\"articleSection\":[\"Agents\",\"Models\",\"Open AI\"],\"inLanguage\":\"en-US\",\"dateModified\":\"2026-11-07T07:24:26+00:00\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents\",\"url\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents\",\"name\":\"BioBasicsBench: OpenAI Tests AI Agents on Wet-Lab Data\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/openai-biobasicsbench-title.webp\",\"datePublished\":\"2026-11-07T07:24:26+00:00\",\"description\":\"BioBasicsBench, built by OpenAI and Ligo, tests AI agents on 100 routine wet-lab tasks. GPT-6 Astra earns 66.7% of credit and still miscounts colonies.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents#primaryimage\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/openai-biobasicsbench-title.webp\",\"contentUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/openai-biobasicsbench-title.webp\",\"width\":1344,\"height\":768,\"caption\":\"Clay-style lab bench with a petri dish, a glowing gel tray and a scientist robot inspecting them through a magnifier\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/openai-biobasicsbench-wet-lab-ai-agents#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/i10x.ai\\\/news\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"OpenAI&#8217;s BioBasicsBench Tests AI Agents on Real Wet-Lab Data\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#website\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\",\"name\":\"i10X News\",\"description\":\"Stay updated with the latest news and announcements from i10X.\",\"publisher\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/i10x.ai\\\/news?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\",\"name\":\"i10X\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/i10x-logo-180.png\",\"contentUrl\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/i10x-logo-180.png\",\"width\":180,\"height\":180,\"caption\":\"i10X\"},\"image\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/x.com\\\/i10x_ai\",\"https:\\\/\\\/www.linkedin.com\\\/company\\\/i10x-ai\",\"https:\\\/\\\/instagram.com\\\/i10x.ai\",\"https:\\\/\\\/www.youtube.com\\\/@i10X_ai_official\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/person\\\/d410bb7ee72febefca2c9c369939219a\",\"name\":\"Christopher Ort\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/christopher-ort-192.webp\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/christopher-ort-192.webp\",\"contentUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/christopher-ort-192.webp\",\"caption\":\"Christopher Ort\"},\"description\":\"Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU). He is the author of \u201cThe Style Penalty\u201d, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/author\\\/christopher-ort\",\"sameAs\":[\"https:\\\/\\\/i10x.ai\\\/blog\\\/author\\\/christopher-ort\"],\"worksFor\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"alumniOf\":{\"@type\":\"EducationalOrganization\",\"name\":\"Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU)\",\"url\":\"https:\\\/\\\/www.fau.eu\"}}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"BioBasicsBench: OpenAI Tests AI Agents on Wet-Lab Data","description":"BioBasicsBench, built by OpenAI and Ligo, tests AI agents on 100 routine wet-lab tasks. GPT-6 Astra earns 66.7% of credit and still miscounts colonies.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents","og_locale":"en_US","og_type":"article","og_title":"BioBasicsBench: OpenAI Tests AI Agents on Wet-Lab Data","og_description":"BioBasicsBench, built by OpenAI and Ligo, tests AI agents on 100 routine wet-lab tasks. GPT-6 Astra earns 66.7% of credit and still miscounts colonies.","og_url":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents","og_site_name":"i10X News","article_published_time":"2026-11-07T07:24:26+00:00","og_image":[{"width":1344,"height":768,"url":"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/11\/openai-biobasicsbench-title.webp","type":"image\/webp"}],"author":"Christopher Ort","twitter_card":"summary_large_image","twitter_creator":"@i10x_ai","twitter_site":"@i10x_ai","twitter_misc":{"Written by":"Christopher Ort","Est. reading time":"8 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents#article","isPartOf":{"@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents"},"author":{"name":"Christopher Ort","@id":"https:\/\/i10x.ai\/news\/#\/schema\/person\/d410bb7ee72febefca2c9c369939219a"},"headline":"OpenAI&#8217;s BioBasicsBench Tests AI Agents on Real Wet-Lab Data","datePublished":"2026-11-07T07:24:26+00:00","mainEntityOfPage":{"@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents"},"wordCount":1619,"publisher":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"image":{"@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/openai-biobasicsbench-title.webp","articleSection":["Agents","Models","Open AI"],"inLanguage":"en-US","dateModified":"2026-11-07T07:24:26+00:00"},{"@type":"WebPage","@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents","url":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents","name":"BioBasicsBench: OpenAI Tests AI Agents on Wet-Lab Data","isPartOf":{"@id":"https:\/\/i10x.ai\/news\/#website"},"primaryImageOfPage":{"@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents#primaryimage"},"image":{"@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/openai-biobasicsbench-title.webp","datePublished":"2026-11-07T07:24:26+00:00","description":"BioBasicsBench, built by OpenAI and Ligo, tests AI agents on 100 routine wet-lab tasks. GPT-6 Astra earns 66.7% of credit and still miscounts colonies.","breadcrumb":{"@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents#primaryimage","url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/openai-biobasicsbench-title.webp","contentUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/openai-biobasicsbench-title.webp","width":1344,"height":768,"caption":"Clay-style lab bench with a petri dish, a glowing gel tray and a scientist robot inspecting them through a magnifier"},{"@type":"BreadcrumbList","@id":"https:\/\/i10xnews.kinsta.cloud\/openai-biobasicsbench-wet-lab-ai-agents#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/i10x.ai\/news"},{"@type":"ListItem","position":2,"name":"OpenAI&#8217;s BioBasicsBench Tests AI Agents on Real Wet-Lab Data"}]},{"@type":"WebSite","@id":"https:\/\/i10x.ai\/news\/#website","url":"https:\/\/i10x.ai\/news","name":"i10X News","description":"Stay updated with the latest news and announcements from i10X.","publisher":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/i10x.ai\/news?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/i10x.ai\/news\/#organization","name":"i10X","url":"https:\/\/i10x.ai\/news\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10x.ai\/news\/#\/schema\/logo\/image\/","url":"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/10\/i10x-logo-180.png","contentUrl":"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/10\/i10x-logo-180.png","width":180,"height":180,"caption":"i10X"},"image":{"@id":"https:\/\/i10x.ai\/news\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/x.com\/i10x_ai","https:\/\/www.linkedin.com\/company\/i10x-ai","https:\/\/instagram.com\/i10x.ai","https:\/\/www.youtube.com\/@i10X_ai_official"]},{"@type":"Person","@id":"https:\/\/i10x.ai\/news\/#\/schema\/person\/d410bb7ee72febefca2c9c369939219a","name":"Christopher Ort","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/christopher-ort-192.webp","url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/christopher-ort-192.webp","contentUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/christopher-ort-192.webp","caption":"Christopher Ort"},"description":"Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU). He is the author of \u201cThe Style Penalty\u201d, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.","url":"https:\/\/i10x.ai\/news\/author\/christopher-ort","sameAs":["https:\/\/i10x.ai\/blog\/author\/christopher-ort"],"worksFor":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"alumniOf":{"@type":"EducationalOrganization","name":"Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU)","url":"https:\/\/www.fau.eu"}}]}},"_links":{"self":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/6974","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/comments?post=6974"}],"version-history":[{"count":0,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/6974\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media\/6973"}],"wp:attachment":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media?parent=6974"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/categories?post=6974"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/tags?post=6974"},{"taxonomy":"company","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/companies?post=6974"},{"taxonomy":"product","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/products?post=6974"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}