[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"post:\u002Fpost\u002Fclaude-opus-47-is-here-best-coding-model-3-vision-same-price":3,"categories-en":502},{"post":4,"related":438,"translations":499,"resolved":17},{"id":5,"slug":6,"title":7,"excerpt":8,"image":9,"featured":10,"groupId":11,"publishedAt":12,"readingTime":13,"views":14,"author":15,"category":18,"tags":22,"contentJson":29,"metaTitle":435,"metaDescription":436,"updatedAt":437},"643964c8-bef9-4e14-8cd6-a7688786d5e6","claude-opus-47-is-here-best-coding-model-3-vision-same-price","Claude Opus 4.7 Is Here: Best Coding Model, 3× Vision, Same Price","Anthropic's Claude Opus 4.7 is here, delivering massive gains in coding and vision at no extra cost. Discover the new features and benchmark results.","https:\u002F\u002Fmfgiomkkbihnlpfdxwrt.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fblog-assets\u002Fmedia\u002Ffeatured\u002F1776422682969-featured-drq31h.jpeg",false,"1f08cd4a-0055-436e-a510-4106236ac465","2026-06-19T21:35:27.596075+00:00",6,0,{"name":16,"avatar":17},"Genildo Souza",null,{"slug":19,"name":20,"color":21},"tech-news","Tech News","#e50606",[23,26],{"slug":24,"name":25},"ai","AI",{"slug":27,"name":28},"claude","Claude",{"type":30,"content":31},"doc",[32,43,51,57,75,83,89,95,107,113,119,130,136,166,172,178,184,195,201,207,212,231,247,253,259,284,295,306,312,318,328,338,348,354,360,366,376,382,394,406,420,426,432],{"type":33,"attrs":34},"summaryBox",{"id":35,"items":36,"title":42},"sum_jdtbfe93",[37,38,39,40,41],"Claude Opus 4.7 delivers significant performance gains in coding benchmarks, achieving 64.3% on SWE-bench Pro and 70% on CursorBench.","The model features a 3x increase in vision resolution, allowing for high-fidelity analysis of complex technical diagrams and IDE screenshots.","Anthropic introduced the xhigh effort level to optimize token usage and improve success rates for complex coding tasks.","New features include task budgets for autonomous agents, an \u002Fultrareview command for code quality, and improved cross-session memory.","Users should note that the new tokenizer may increase token counts by up to 35% and that prompts may require adjustments due to stricter instruction following.","Claude Opus 4.7: Key Features and Upgrades",{"type":44,"attrs":45,"content":47},"paragraph",{"id":46},"p_ygh5o4yj",[48],{"text":49,"type":50},"Struggling with AI models that stumble on complex coding tasks or lack the vision to handle your most demanding visual data? Anthropic has just released Claude Opus 4.7 to solve exactly that. This upgrade delivers their most powerful coding performance yet and triples your vision capabilities—all while keeping the price completely unchanged. It is available now on claude.ai, the API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.","text",{"type":44,"attrs":52,"content":54},{"id":53},"p_ggnpjfre",[55],{"text":56,"type":50},"This is not an incremental update. SWE-bench Pro is up 10.9 percentage points. CursorBench is up 12 points. Vision resolution has tripled. And — a detail companies are happy to hear — the pricing has not changed.",{"type":58,"attrs":59},"archVisual",{"id":60,"cards":61,"title":17,"caption":17},"arch_t0b0w36c",[62,66,71],{"tag":63,"title":64,"subtitle":65},"SWE-bench Pro","64.3%","era 53,4% no Opus 4.6 (+10,9pp)",{"tag":67,"title":68,"accent":69,"subtitle":70},"CursorBench","70%",1,"era 58% (+12pp) — melhor coding do mercado",{"tag":72,"title":73,"accent":69,"subtitle":74},"Visão","3.75MP","era 1,15MP — 3× mais resolução",{"type":76,"attrs":77,"content":80},"heading",{"id":78,"level":79},"h_bl4ztwhm",2,[81],{"text":82,"type":50},"What has actually changed",{"type":44,"attrs":84,"content":86},{"id":85},"p_6sekj1tu",[87],{"text":88,"type":50},"Opus 4.7 was built around three real problems that Opus 4.6 users were reporting: the model sometimes abandoned long tasks midway, sometimes delivered code that looked correct but failed review, and sometimes interpreted instructions more loosely than expected.",{"type":44,"attrs":90,"content":92},{"id":91},"p_ep4gqtfl",[93],{"text":94,"type":50},"The three central bets of Opus 4.7 are directly aimed at these problems: persistence in long tasks, self-verification before reporting, and literal instruction following.",{"type":96,"attrs":97,"content":100},"callout",{"id":98,"type":99,"title":17},"call_aqm9w68d","info",[101],{"type":44,"attrs":102,"content":104},{"id":103},"p_3tx3wvwa",[105],{"text":106,"type":50},"Opus 4.7 at low effort is equivalent to Opus 4.6 at medium effort — in practice, you pay the same and get more quality per token.",{"type":76,"attrs":108,"content":110},{"id":109,"level":79},"h_c78665p7",[111],{"text":112,"type":50},"Benchmarks: where Opus 4.7 won and where it gave ground",{"type":114,"attrs":115},"svgDiagram",{"id":116,"svg":117,"label":118},"svg_c252dw1y","\u003Csvg xmlns=\"http:\u002F\u002Fwww.w3.org\u002F2000\u002Fsvg\" viewBox=\"0 0 680 260\" font-family=\"'Inter','Segoe UI',sans-serif\">\n  \u003Ctext x=\"40\" y=\"22\" font-size=\"12\" fill=\"#64748b\" font-weight=\"700\">Opus 4.6 vs Opus 4.7 — benchmarks principais\u003C\u002Ftext>\n  \u003Ctext x=\"510\" y=\"22\" font-size=\"12\" fill=\"#94a3b8\">■ Opus 4.6  ■ Opus 4.7\u003C\u002Ftext>\n  \u003C!-- SWE-bench Verified -->\n  \u003Ctext x=\"40\" y=\"50\" font-size=\"12\" fill=\"#64748b\">SWE-bench Verified\u003C\u002Ftext>\n  \u003Crect x=\"220\" y=\"38\" width=\"242\" height=\"14\" rx=\"2\" fill=\"rgba(100,116,139,0.3)\"\u002F>\n  \u003Crect x=\"220\" y=\"38\" width=\"263\" height=\"14\" rx=\"2\" fill=\"rgba(249,115,22,0.7)\"\u002F>\n  \u003Ctext x=\"492\" y=\"50\" font-size=\"12\" fill=\"#f97316\" font-weight=\"700\">87,6%\u003C\u002Ftext>\n  \u003Ctext x=\"555\" y=\"50\" font-size=\"12\" fill=\"#94a3b8\">(+6,8pp)\u003C\u002Ftext>\n  \u003C!-- SWE-bench Pro -->\n  \u003Ctext x=\"40\" y=\"80\" font-size=\"12\" fill=\"#64748b\">SWE-bench Pro\u003C\u002Ftext>\n  \u003Crect x=\"220\" y=\"68\" width=\"160\" height=\"14\" rx=\"2\" fill=\"rgba(100,116,139,0.3)\"\u002F>\n  \u003Crect x=\"220\" y=\"68\" width=\"193\" height=\"14\" rx=\"2\" fill=\"rgba(249,115,22,0.7)\"\u002F>\n  \u003Ctext x=\"422\" y=\"80\" font-size=\"12\" fill=\"#f97316\" font-weight=\"700\">64,3%\u003C\u002Ftext>\n  \u003Ctext x=\"480\" y=\"80\" font-size=\"12\" fill=\"#94a3b8\">(era 53,4% — +10,9pp)\u003C\u002Ftext>\n  \u003C!-- CursorBench -->\n  \u003Ctext x=\"40\" y=\"110\" font-size=\"12\" fill=\"#64748b\">CursorBench\u003C\u002Ftext>\n  \u003Crect x=\"220\" y=\"98\" width=\"174\" height=\"14\" rx=\"2\" fill=\"rgba(100,116,139,0.3)\"\u002F>\n  \u003Crect x=\"220\" y=\"98\" width=\"210\" height=\"14\" rx=\"2\" fill=\"rgba(249,115,22,0.7)\"\u002F>\n  \u003Ctext x=\"439\" y=\"110\" font-size=\"12\" fill=\"#f97316\" font-weight=\"700\">70%\u003C\u002Ftext>\n  \u003Ctext x=\"470\" y=\"110\" font-size=\"12\" fill=\"#94a3b8\">(era 58% — +12pp)\u003C\u002Ftext>\n  \u003C!-- GPQA Diamond -->\n  \u003Ctext x=\"40\" y=\"140\" font-size=\"12\" fill=\"#64748b\">GPQA Diamond\u003C\u002Ftext>\n  \u003Crect x=\"220\" y=\"128\" width=\"274\" height=\"14\" rx=\"2\" fill=\"rgba(100,116,139,0.3)\"\u002F>\n  \u003Crect x=\"220\" y=\"128\" width=\"283\" height=\"14\" rx=\"2\" fill=\"rgba(249,115,22,0.7)\"\u002F>\n  \u003Ctext x=\"512\" y=\"140\" font-size=\"12\" fill=\"#f97316\" font-weight=\"700\">94,2%\u003C\u002Ftext>\n  \u003Ctext x=\"560\" y=\"140\" font-size=\"12\" fill=\"#94a3b8\">(+2,9pp)\u003C\u002Ftext>\n  \u003C!-- Finance Agent -->\n  \u003Ctext x=\"40\" y=\"170\" font-size=\"12\" fill=\"#64748b\">Finance Agent v1.1\u003C\u002Ftext>\n  \u003Crect x=\"220\" y=\"158\" width=\"182\" height=\"14\" rx=\"2\" fill=\"rgba(100,116,139,0.3)\"\u002F>\n  \u003Crect x=\"220\" y=\"158\" width=\"193\" height=\"14\" rx=\"2\" fill=\"rgba(249,115,22,0.7)\"\u002F>\n  \u003Ctext x=\"422\" y=\"170\" font-size=\"12\" fill=\"#f97316\" font-weight=\"700\">64,4%\u003C\u002Ftext>\n  \u003Ctext x=\"470\" y=\"170\" font-size=\"12\" fill=\"#94a3b8\">(era 60,7% — melhor do mercado)\u003C\u002Ftext>\n  \u003C!-- BrowseComp REGRESSÃO -->\n  \u003Ctext x=\"40\" y=\"200\" font-size=\"12\" fill=\"#64748b\">BrowseComp\u003C\u002Ftext>\n  \u003Crect x=\"220\" y=\"188\" width=\"251\" height=\"14\" rx=\"2\" fill=\"rgba(249,115,22,0.5)\"\u002F>\n  \u003Crect x=\"220\" y=\"188\" width=\"238\" height=\"14\" rx=\"2\" fill=\"rgba(239,68,68,0.6)\"\u002F>\n  \u003Ctext x=\"467\" y=\"200\" font-size=\"12\" fill=\"#ef4444\" font-weight=\"700\">79,3%\u003C\u002Ftext>\n  \u003Ctext x=\"515\" y=\"200\" font-size=\"12\" fill=\"#ef4444\">(era 83,7% — regressão)\u003C\u002Ftext>\n  \u003Ctext x=\"40\" y=\"240\" font-size=\"12\" fill=\"#64748b\" font-style=\"italic\">Barras laranjas = Opus 4.7 · Barras com borda vermelha = regressão vs 4.6\u003C\u002Ftext>\n\u003C\u002Fsvg>","Opus 4.6 vs Opus 4.7 comparison on key benchmarks",{"type":96,"attrs":120,"content":123},{"id":121,"type":122,"title":17},"call_liyjwv0o","warning",[124],{"type":44,"attrs":125,"content":127},{"id":126},"p_5xkc0yx9",[128],{"text":129,"type":50},"BrowseComp dropped from 83.7% to 79.3% — a real regression in multi-page web research. If your agents rely heavily on internet search and synthesis, GPT-5.4 Pro (89.3%) still holds an advantage in this specific benchmark.",{"type":76,"attrs":131,"content":133},{"id":132,"level":79},"h_tenzto16",[134],{"text":135,"type":50},"Opus 4.7 vs GPT-5.4 vs Gemini 3.1 Pro",{"type":137,"attrs":138,"content":140},"compare",{"id":139},"cmp_tqnohcq0",[141,154],{"type":142,"attrs":143,"content":147},"compareColumn",{"id":144,"side":145,"title":146},"col_oi8pgs0g","left","Claude Opus 4.7 leads",[148],{"type":44,"attrs":149,"content":151},{"id":150},"p_wmbt651y",[152],{"text":153,"type":50},"SWE-bench Pro: 64.3% vs 57.7% (GPT) and 54.2% (Gemini)\n\nSWE-bench Verified: 87.6%\n\nCursorBench: 70% — best coding in IDE on the market\n\nMCP-Atlas (tool use): 77.3% vs 68.1% (GPT)\n\nFinance Agent: 64.4% vs 59.7% (Gemini)\n\nGDPVal-AA knowledge work: Elo 1,753 vs 1,674 (GPT)",{"type":142,"attrs":155,"content":159},{"id":156,"side":157,"title":158},"col_fvpgpet8","right","Where it loses or ties",[160],{"type":44,"attrs":161,"content":163},{"id":162},"p_03dw4cxv",[164],{"text":165,"type":50},"BrowseComp: 79.3% vs 89.3% (GPT) and 85.9% (Gemini)\n\nGPQA Diamond: 94.2% — practically tied (GPT: 94.4%, Gemini: 94.3%)\n\nTerminal-Bench 2.0: 69.4% vs 75.1% (GPT)\n\nHumanity's Last Exam: 54.7% vs 58.7% (GPT)\n\nCyberGym: intentional — cyber capabilities were reduced during training",{"type":76,"attrs":167,"content":169},{"id":168,"level":79},"h_8kx13l18",[170],{"text":171,"type":50},"3× better vision — what this changes in practice",{"type":44,"attrs":173,"content":175},{"id":174},"p_9620n84v",[176],{"text":177,"type":50},"Opus 4.6 processed images at up to 1,568px on the long side (1.15 megapixels). Opus 4.7 goes up to 2,576px (3.75 megapixels) — more than 3× more pixels.",{"type":44,"attrs":179,"content":181},{"id":180},"p_j6ejvh11",[182],{"text":183,"type":50},"In practice: dense technical diagrams, IDE screenshots, high-resolution PDF documents, design mockups, and complex financial charts arrive with true fidelity — not interpolated. The CharXiv visual reasoning with tools benchmark jumped from 84.7% to 91.0%.",{"type":96,"attrs":185,"content":188},{"id":186,"type":187,"title":17},"call_kmd7o07j","tip",[189],{"type":44,"attrs":190,"content":192},{"id":191},"p_zua5lmtn",[193],{"text":194,"type":50},"Computer use agents will feel this difference in interface navigation. OSWorld-Verified rose to 78.0%. For analysis of scientific patents and chemical diagrams, partners report significant improvements.",{"type":76,"attrs":196,"content":198},{"id":197,"level":79},"h_d6t6w80g",[199],{"text":200,"type":50},"The new xhigh level — fine control between quality and cost",{"type":44,"attrs":202,"content":204},{"id":203},"p_zianwxiw",[205],{"text":206,"type":50},"Opus 4.6 had four effort levels: low, medium, high, and max. Opus 4.7 introduces a new level between high and max:",{"type":114,"attrs":208},{"id":209,"svg":210,"label":211},"svg_fvdv564j","\u003Csvg xmlns=\"http:\u002F\u002Fwww.w3.org\u002F2000\u002Fsvg\" viewBox=\"0 0 680 80\" font-family=\"'Inter','Segoe UI',sans-serif\">\n  \u003Cline x1=\"40\" y1=\"40\" x2=\"640\" y2=\"40\" stroke=\"#1e2330\" stroke-width=\"2\"\u002F>\n  \u003Ccircle cx=\"80\" cy=\"40\" r=\"8\" fill=\"#3b82f6\"\u002F>\n  \u003Ctext x=\"80\" y=\"22\" text-anchor=\"middle\" font-size=\"12\" fill=\"#3b82f6\">low\u003C\u002Ftext>\n  \u003Ctext x=\"80\" y=\"62\" text-anchor=\"middle\" font-size=\"12\" fill=\"#64748b\">econômico\u003C\u002Ftext>\n  \u003Ccircle cx=\"213\" cy=\"40\" r=\"8\" fill=\"#3b82f6\"\u002F>\n  \u003Ctext x=\"213\" y=\"22\" text-anchor=\"middle\" font-size=\"12\" fill=\"#3b82f6\">medium\u003C\u002Ftext>\n  \u003Ctext x=\"213\" y=\"62\" text-anchor=\"middle\" font-size=\"12\" fill=\"#64748b\">balanceado\u003C\u002Ftext>\n  \u003Ccircle cx=\"346\" cy=\"40\" r=\"8\" fill=\"#3b82f6\"\u002F>\n  \u003Ctext x=\"346\" y=\"22\" text-anchor=\"middle\" font-size=\"12\" fill=\"#3b82f6\">high\u003C\u002Ftext>\n  \u003Ctext x=\"346\" y=\"62\" text-anchor=\"middle\" font-size=\"12\" fill=\"#64748b\">padrão Claude Code\u003C\u002Ftext>\n  \u003Ccircle cx=\"480\" cy=\"40\" r=\"12\" fill=\"#f97316\"\u002F>\n  \u003Ctext x=\"480\" y=\"22\" text-anchor=\"middle\" font-size=\"12\" fill=\"#f97316\" font-weight=\"700\">xhigh ✦ novo\u003C\u002Ftext>\n  \u003Ctext x=\"480\" y=\"65\" text-anchor=\"middle\" font-size=\"12\" fill=\"#f97316\">novo padrão Claude Code\u003C\u002Ftext>\n  \u003Ccircle cx=\"600\" cy=\"40\" r=\"8\" fill=\"#94a3b8\"\u002F>\n  \u003Ctext x=\"600\" y=\"22\" text-anchor=\"middle\" font-size=\"12\" fill=\"#94a3b8\">max\u003C\u002Ftext>\n  \u003Ctext x=\"600\" y=\"62\" text-anchor=\"middle\" font-size=\"12\" fill=\"#64748b\">máximo — caro\u003C\u002Ftext>\n\u003C\u002Fsvg>","Effort level scale in Opus 4.7",{"type":44,"attrs":213,"content":215},{"id":214},"p_cla1wjhb",[216,218,223,225,229],{"text":217,"type":50},"xhigh is now the default for Claude Code across all plans. The logic is simple: if a task requires three attempts at ",{"text":219,"type":50,"marks":220},"high",[221],{"type":222},"code",{"text":224,"type":50}," to get it right, one attempt at ",{"text":226,"type":50,"marks":227},"xhigh",[228],{"type":222},{"text":230,"type":50}," is usually cheaper in total — fewer retries, fewer tokens spent.",{"type":96,"attrs":232,"content":234},{"id":233,"type":99,"title":17},"call_xkh22api",[235],{"type":44,"attrs":236,"content":238},{"id":237},"p_c50yhc4h",[239,241,245],{"text":240,"type":50},"In the API: use ",{"text":242,"type":50,"marks":243},"thinking: {type: \"adaptive\"}",[244],{"type":222},{"text":246,"type":50}," with the effort parameter. Extended manual thinking has been removed in Opus 4.7 — it is now exclusively adaptive.",{"type":76,"attrs":248,"content":250},{"id":249,"level":79},"h_h15s38tv",[251],{"text":252,"type":50},"Task budgets, \u002Fultrareview and cross-session memory",{"type":44,"attrs":254,"content":256},{"id":255},"p_cd0h5nos",[257],{"text":258,"type":50},"Three new features arriving with the model:",{"type":44,"attrs":260,"content":262},{"id":261},"p_gbtmhhmd",[263,265,270,272,276,278,282],{"text":264,"type":50},"— ",{"text":266,"type":50,"marks":267},"Task budgets (public beta)",[268],{"type":269},"bold",{"text":271,"type":50},": set a token ceiling for autonomous agents. The model sees the counter decreasing and prioritizes the work, finishing cleanly instead of cutting off abruptly. Activate via header ",{"text":273,"type":50,"marks":274},"task-budgets-2026-03-13",[275],{"type":222},{"text":277,"type":50}," + parameter ",{"text":279,"type":50,"marks":280},"output_config.task_budget",[281],{"type":222},{"text":283,"type":50},".",{"type":44,"attrs":285,"content":287},{"id":286},"p_6m2ebonc",[288,289,293],{"text":264,"type":50},{"text":290,"type":50,"marks":291},"\u002Fultrareview in Claude Code",[292],{"type":269},{"text":294,"type":50},": new command that runs a dedicated review session, reads the entire diff, and flags what a careful human reviewer would detect. 3 free uses on Pro and Max plans at launch.",{"type":44,"attrs":296,"content":298},{"id":297},"p_8g1xydxf",[299,300,304],{"text":264,"type":50},{"text":301,"type":50,"marks":302},"Cross-session memory",[303],{"type":269},{"text":305,"type":50},": Opus 4.7 is better at using file-system-based memory. It keeps important notes between long work sessions, reducing the context you need to paste at the start of each new session.",{"type":76,"attrs":307,"content":309},{"id":308,"level":79},"h_6y0wq42u",[310],{"text":311,"type":50},"Attention on the 4.6 migration",{"type":44,"attrs":313,"content":315},{"id":314},"p_s7nx7gc4",[316],{"text":317,"type":50},"Anthropic called it a \"direct upgrade\" but there are changes that affect token usage and behavior:",{"type":96,"attrs":319,"content":321},{"id":320,"type":122,"title":17},"call_gxnzhf16",[322],{"type":44,"attrs":323,"content":325},{"id":324},"p_i37jzx46",[326],{"text":327,"type":50},"1. New tokenizer: the same input can generate 1.0 to 1.35× more tokens. Measure in real traffic before migrating production.",{"type":96,"attrs":329,"content":331},{"id":330,"type":122,"title":17},"call_fxa4febq",[332],{"type":44,"attrs":333,"content":335},{"id":334},"p_jj4q9i11",[336],{"text":337,"type":50},"2. Literal instructions: Opus 4.7 follows instructions more precisely. Prompts tuned for the looser behavior of Opus 4.6 may need revision.",{"type":96,"attrs":339,"content":341},{"id":340,"type":122,"title":17},"call_g9bnoym3",[342],{"type":44,"attrs":343,"content":345},{"id":344},"p_h4o15ppe",[346],{"text":347,"type":50},"3. Extended thinking budgets removed — use adaptive thinking. Temperature\u002Ftop_p\u002Ftop_k parameters were also removed — control via prompting.",{"type":76,"attrs":349,"content":351},{"id":350,"level":79},"h_0vsmp2wt",[352],{"text":353,"type":50},"The elephant in the room: the Mythos Preview",{"type":44,"attrs":355,"content":357},{"id":356},"p_pitjzvud",[358],{"text":359,"type":50},"Anthropic was transparent: Opus 4.7 does not match the Claude Mythos Preview, their most powerful model — which is not publicly available due to safety concerns.",{"type":44,"attrs":361,"content":363},{"id":362},"p_q5dobaag",[364],{"text":365,"type":50},"The Mythos Preview was released last week to a select group of technology and cybersecurity companies as part of Project Glasswing. Opus 4.7 is the first model where Anthropic tested safeguards against use in cyberattacks — what they learn here will guide how they eventually release Mythos-level models at scale.",{"type":96,"attrs":367,"content":369},{"id":368,"type":99,"title":17},"call_xab2kypz",[370],{"type":44,"attrs":371,"content":373},{"id":372},"p_c7vhqieg",[374],{"text":375,"type":50},"Security professionals who want to use Opus 4.7 for vulnerability research, penetration testing, and red-teaming can sign up for Anthropic's Cyber Verification Program.",{"type":76,"attrs":377,"content":379},{"id":378,"level":79},"h_rozp9qms",[380],{"text":381,"type":50},"Pricing, availability, and model ID",{"type":44,"attrs":383,"content":385},{"id":384},"p_l8luo93q",[386,388,392],{"text":387,"type":50},"Price identical to Opus 4.6: ",{"text":389,"type":50,"marks":390},"$5 per million input tokens and $25 per million output tokens",[391],{"type":269},{"text":393,"type":50},". Prompt caching reduces by up to 90%. Batch processing reduces by 50%.",{"type":44,"attrs":395,"content":397},{"id":396},"p_utvfd0ch",[398,400,404],{"text":399,"type":50},"Model ID in the API: ",{"text":401,"type":50,"marks":402},"claude-opus-4-7",[403],{"type":222},{"text":405,"type":50},". Available on: claude.ai (Pro, Max, Team, Enterprise), Anthropic API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.",{"type":33,"attrs":407},{"id":408,"items":409,"title":419},"sum_ujthof49",[410,411,412,413,414,415,416,417,418],"SWE-bench Pro: 64.3% (+10.9pp vs 4.6) — best coding model generally available on the market today.","CursorBench: 70% (+12pp) — the benchmark that measures what developers actually do day-to-day in an IDE.","3.75MP vision (was 1.15MP) — technical diagrams, screenshots, and mockups arrive with true fidelity.","xhigh: new effort level between high and max — now default in Claude Code. More quality without paying the max cost.","Task budgets (beta): token ceiling for autonomous agents. No more billing surprises on overnight runs.","\u002Fultrareview in Claude Code: multi-agent parallel diff review with bug and design issue flagging.","Same price as Opus 4.6: $5\u002F$25 per million tokens. There is no cheaper free upgrade than this.","BrowseComp regressed: 79.3% (was 83.7%). Agents that rely on intensive web research should test before migrating.","Breaking changes in migration: new tokenizer (1.0–1.35× more tokens), more literal instructions, mandatory adaptive thinking.","What remains of Opus 4.7",{"type":44,"attrs":421,"content":423},{"id":422},"p_uu17vfbe",[424],{"text":425,"type":50},"For most developers using Claude Code day-to-day, Opus 4.7 is a direct upgrade with no decision to make. Same price, better model.",{"type":44,"attrs":427,"content":429},{"id":428},"p_raipl5yd",[430],{"text":431,"type":50},"For teams with agents in production, the migration requires attention: measure the impact of the new tokenizer, review prompts that relied on loose interpretation, and configure task budgets before turning on auto mode.",{"type":44,"attrs":433},{"id":434},"p_ouu710hz","Claude Opus 4.7: Best Coding Model, 3x Vision, Same Price","Explore Claude Opus 4.7: featuring 3x higher vision resolution, improved coding benchmarks, new task budgets, and the efficient xhigh effort level.","2026-08-24T14:53:53.315534+00:00",[439,458,477],{"id":440,"slug":441,"title":442,"excerpt":443,"image":444,"featured":10,"groupId":445,"publishedAt":446,"readingTime":447,"views":14,"author":448,"category":449,"tags":450},"036e0cc5-1c1a-483a-8c6b-3e384f6cb5f7","copilot-cli-becomes-an-agent-writes-code-runs-tests-and-fixes-bugs-without-prompting","Copilot CLI Becomes an Agent: Writes Code, Runs Tests, and Fixes Bugs Without Prompting","GitHub Copilot CLI is now generally available with Autopilot mode, specialized agents, and cross-session memory to automate your terminal workflow.","https:\u002F\u002Fmfgiomkkbihnlpfdxwrt.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fblog-assets\u002Fmedia\u002Ffeatured\u002F1776233080837-featured-igzqoq.jpg","6c42478b-bd8e-435b-8988-9916cf2f7d6a","2026-07-10T06:17:45.139252+00:00",3,{"name":16,"avatar":17},{"slug":19,"name":20,"color":21},[451,452,455],{"slug":24,"name":25},{"slug":453,"name":454},"copilot","Copilot",{"slug":456,"name":457},"microsoft","Microsoft",{"id":459,"slug":460,"title":461,"excerpt":462,"image":463,"featured":10,"groupId":464,"publishedAt":465,"readingTime":466,"views":14,"author":467,"category":468,"tags":469},"3f2f0a1b-78d3-4a87-84e4-f06053ebd9a4","end-of-claude-agent-sdk-subsidies-prepare-your-budget-for-a-15x-to-30x-ai-cost-increase","End of Claude Agent SDK subsidies: prepare your budget for a 15x to 30x AI cost increase","Anthropic has paused the Claude Agent SDK pricing change. Understand the background of this decision and the future of automated agent usage.","https:\u002F\u002Fmfgiomkkbihnlpfdxwrt.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fblog-assets\u002Fmedia\u002Ffeatured\u002F1783133158909-featured-m3trm.webp","b3e08cfe-6221-401e-aced-1fa59a3994d2","2026-06-30T02:21:52.132905+00:00",14,{"name":16,"avatar":17},{"slug":19,"name":20,"color":21},[470,471,474],{"slug":24,"name":25},{"slug":472,"name":473},"developer-tools","Developer Tools",{"slug":475,"name":476},"llm","LLM",{"id":478,"slug":479,"title":480,"excerpt":481,"image":482,"featured":10,"groupId":483,"publishedAt":484,"readingTime":447,"views":485,"author":486,"category":487,"tags":488},"b04ec741-41e2-45d2-b388-043b545c579d","china-leads-the-open-source-ai-race-but-the-foundation-is-still-american","China leads the open-source AI race — but the foundation is still American","The advancement of Chinese open-source models in adoption and downloads shows a significant geopolitical shift in AI. But this victory has a structural limit: much of the compute, tooling, and acceleration infrastructure still relies on a base dominated by a United States company.","https:\u002F\u002Fmfgiomkkbihnlpfdxwrt.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fblog-assets\u002Fmedia\u002Ffeatured\u002F1774414575081-featured-xv6ljv.jpg","7cb14280-9b0b-4afc-baba-5f51ae5aa59e","2026-03-25T19:11:06.877+00:00",106,{"name":16,"avatar":17},{"slug":19,"name":20,"color":21},[489,492,495,498],{"slug":490,"name":491},"open-source","Open Source",{"slug":493,"name":494},"china","China",{"slug":496,"name":497},"nvidia","Nvidia",{"slug":24,"name":25},{"pt":500,"en":6,"es":501},"claude-opus-4-7-lancamento-benchmarks-xhigh","llego-claude-opus-47-mejor-modelo-de-coding-vision-3-mayor-y-mismo-precio",[503,509,515,521,524],{"id":504,"slug":505,"name":506,"colorHex":507,"iconName":17,"postCount":508},"0e2c1225-f78a-4c20-8c93-998ab9dc93b9","frontend","Frontend","#40484f",25,{"id":510,"slug":511,"name":512,"colorHex":513,"iconName":17,"postCount":514},"bc8df70e-b7fb-4d89-a80c-fd458d36d58c","code-development","Code & Development","#8b5cf6",23,{"id":516,"slug":517,"name":518,"colorHex":519,"iconName":520,"postCount":466},"550e8400-e29b-41d4-a716-446655440001","inteligencia-artificial","Artificial Intelligence","#6366f1","brain",{"id":522,"slug":19,"name":20,"colorHex":21,"iconName":17,"postCount":523},"ea63b69c-2eb4-4100-81f3-70ce52d33743",4,{"id":525,"slug":526,"name":527,"colorHex":528,"iconName":529,"postCount":79},"550e8400-e29b-41d4-a716-446655440017","how-to-guides","How-To & Guides","#be123c","book-open"]