Bài học chính

• Nếu chốt nhanh: Claude Opus 4.7 hiện có lợi thế rõ hơn trên các benchmark công khai về reasoning/coding-agent, còn GPT-5.5 được OpenAI định vị là model reasoning mạnh nhất trong ChatGPT cho công việc khó và workflow nhiều bước.[2][10][15] Nhưng để nói model nào “mạnh hơn toàn diệ
• Anthropic mô tả Claude Opus 4.7 là model generally available mạnh nhất của họ cho complex reasoning và agentic coding.[2]

Câu trả lời nghiên cứu

Hỗ trợ hình ảnh

Trao đổi nghiên cứu

Người ta cũng hỏi

Tiếp tục nghiên cứu của bạn

Nguồn

• [1] Claude Platform - Claude API Docsdocs.anthropic.com

  April 16, 2026 We've launched Claude Opus 4.7, our most capable generally available model for complex reasoning and agentic coding, at the same $5 / $25 per MTok pricing as Opus 4.6. See What's new in Claude Opus 4.7 for capability improvements, new features, and the updated tokenizer. Opus 4.7 includes API breaking changes versus Opus 4.6; see Migrating to Claude Opus 4.7 before upgrading. Claude in Amazon Bedrock is now open to all Amazon Bedrock customers. Claude Opus 4.7 and Claude Haiku 4.5 are available self-serve from the Bedrock console through the Messages API endpoint at `/anthr…

• [2] Release notes | Claude Help Centerdocs.anthropic.com

  February 12, 2026 Self-serve Enterprise plans Previously, Enterprise plans were only available to customers working with our Sales team. Now, any organization can purchase an Enterprise plan directly on our website with no Sales conversation required. Self-serve Enterprise plans have a single seat type that includes access to Claude, Claude Code, and Cowork. For more information, refer to our blog post or What is the Enterprise plan? ### February 5, 2026 Claude Opus 4.6 launch We’ve upgraded our smartest model and improved its coding skills. Read our blog post for more information: Introd…

• [3] System Prompts - Claude API Docsdocs.anthropic.com

  System Prompts - Claude API Docs Loading... and mobile apps use a system prompt to provide up-to-date information, such as the current date, to Claude at the start of every conversation. The system prompt also encourages certain behaviors, such as always providing code snippets in Markdown. This prompt is periodically updated to improve Claude's responses. These system prompt updates do not apply to the Claude API. Updates between versions are bolded. ## Claude Opus 4.7 ### April 16, 2026 ## Claude Sonnet 4.6 ### February 17, 2026 ## Claude Opus 4.6 ### February 5, 2026 ## Claude Opus 4.5 #…

• [4] An update on recent Claude Code quality reports - Anthropicanthropic.com

  A few weeks before we released Opus 4.7, we started tuning Claude Code in preparation. Each model behaves slightly differently, and we spend time before each release optimizing the harness and product for it. We have a number of tools to reduce verbosity: model training, prompting, and improving thinking UX in the product. Ultimately we used all of these, but one addition to the system prompt caused an outsized effect on intelligence in Claude Code: > “Length limits: keep text between tool calls to ≤25 words. Keep final responses to ≤100 words unless the task requires more detail.” After mu…

• [5] Anthropic expands partnership with Google and Broadcom for ...anthropic.com

  Skip to main contentSkip to footer , Google Cloud (Vertex AI), and Microsoft Azure (Foundry). []( ## Related content ### Anthropic and NEC collaborate to build Japan’s largest AI engineering workforce Read more ### Introducing Claude Design by Anthropic Labs Today, we’re launching Claude Design, a new Anthropic Labs product that lets you collaborate with Claude to create polished visual work like designs, prototypes, slides, one-pagers, and more. Read more ### Introducing Claude Opus 4.7 Our latest Opus model brings stronger performance across coding, agents, vision, and multi-step tasks, wit…

• [6] How up-to-date is Claude's training data? | Claude Help Centersupport.anthropic.com

  Claude Help Center # How up-to-date is Claude's training data? While we're constantly updating Claude's data, each model has a knowledge cutoff: Claude Opus 4.7 was trained on data up until January 2026. Claude Sonnet 4.6 was trained on data up until August 2025. Claude Opus 4.6 was trained on data up until August 2025. Claude Opus 4.5 was trained on data up until August 2025. Claude Haiku 4.5 was trained on data up until July 2025. Claude Sonnet 4.5 was trained on data up until July 2025. Claude Opus 3 was trained on data up until August 2023. These models may not be aware of events or infor…

• [7] Introducing Claude Design by Anthropic Labsanthropic.com

  For Enterprise organizations, Claude Design is off by default. Admins can enable it in Organization settings. Start designing at claude.ai/design. []( ## Related content ### Anthropic and NEC collaborate to build Japan’s largest AI engineering workforce Read more ### Introducing Claude Opus 4.7 Our latest Opus model brings stronger performance across coding, agents, vision, and multi-step tasks, with greater thoroughness and consistency on the work that matters most. Read more ### Anthropic’s Long-Term Benefit Trust appoints Vas Narasimhan to Board of Directors Read more []( ### Products [...…

• [8] Introducing Claude Opus 4.7 - Anthropicanthropic.com

  Migrating from Opus 4.6 to Opus 4.7 Opus 4.7 is a direct upgrade to Opus 4.6, but two changes are worth planning for because they affect token usage. First, Opus 4.7 uses an updated tokenizer that improves how the model processes text. The tradeoff is that the same input can map to more tokens—roughly 1.0–1.35× depending on the content type. Second, Opus 4.7 thinks more at higher effort levels, particularly on later turns in agentic settings. This improves its reliability on hard problems, but it does mean it produces more output tokens. [...] Image 24: logo > Claude Opus 4.7 is a solid up…

• [9] Model system cards - Anthropicanthropic.com

  | Model | Date | System card | --- | Claude Opus 4.7 | April 2026 | Read system card | | Mythos Preview | April 2026 | Read system card | | Claude Sonnet 4.6 | February 2026 | Read system card | | Claude Opus 4.6 | February 2026 | Read system card | | Claude Opus 4.5 | November 2025 | Read system card | | Claude Haiku 4.5 | October 2025 | Read system card | | Claude Sonnet 4.5 | September 2025 | Read system card | | Claude Opus 4.1 | August 2025 | Read system card | | Claude Sonnet 4 and Opus 4 | May 2025 | Read system card | | Claude Sonnet 3.7 | February 2025 | Read system card | | Claude H…

• [10] Newsroom - Anthropicanthropic.com

  Apr 24, 2026 Announcements Anthropic and NEC collaborate to build Japan’s largest AI engineering workforce Apr 17, 2026 Product Introducing Claude Design by Anthropic Labs Apr 16, 2026 Product Introducing Claude Opus 4.7 Apr 14, 2026 Announcements Anthropic’s Long-Term Benefit Trust appoints Vas Narasimhan to Board of Directors Apr 6, 2026 Announcements Anthropic expands partnership with Google and Broadcom for multiple gigawatts of next-generation compute Mar 31, 2026 Announcements Australian government and Anthropic sign MOU for AI safety and research Mar 12, 2026 Announcements Anthropic in…

• [11] Research - Anthropicanthropic.com

  Apr 22, 2026 Economic Research Announcing the Anthropic Economic Index Survey Apr 22, 2026 Economic Research What 81,000 people told us about the economics of AI Apr 14, 2026 Alignment Automated Alignment Researchers: Using large language models to scale scalable oversight Apr 9, 2026 Policy Trustworthy agents in practice Apr 2, 2026 Interpretability Emotion concepts and their function in a large language model Mar 31, 2026 Economic Research How Australia Uses Claude: Findings from the Anthropic Economic Index Mar 24, 2026 Economic Research Anthropic Economic Index report: Learning curves Mar…

• [12] Home \ Anthropicanthropic.com

  Model detailsModel details Model details Date April 16, 2026 Category Announcements Read announcementRead announcement Read announcement ### Claude is a space to think No ads. No sponsored content. Just genuinely helpful conversations. Date February 4, 2026 Category Announcements Read the postRead the post Read the post ### Claude on Mars The first AI-assisted drive on another planet. Claude helped NASA’s Perseverance rover travel four hundred meters on Mars. Date January 30, 2026 Category Announcements Read the storyRead the story Read the story ## At Anthropic, we build AI to serve humanity…

• [13] Codex changelog - OpenAI Developersdevelopers.openai.com

  Changelog Feature Maturity Open Source April 2026 March 2026 February 2026 January 2026 December 2025 November 2025 October 2025 September 2025 August 2025 June 2025 May 2025 # Codex changelog Latest updates to Codex, OpenAI’s coding agent All updatesGeneralCodex appCodex CLI April 2026March 2026February 2026January 2026December 2025November 2025October 2025September 2025August 2025June 2025May 2025 ## April 2026 2026-04-23 ### GPT-5.5 and Codex app updates GPT-5.5 is now available in Codex as OpenAI’s newest frontier model for complex coding, computer use, knowledge work, and research workfl…

• [14] Announcements - OpenAI Developer Communitycommunity.openai.com

  Topic list, column headers with buttons are sortable.| Topic | Posters | Replies | Views | Activity | --- --- | GPT-5.5 is here! Available in Codex and ChatGPT today models Introducing GPT-5.5 A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting …read more | Image 2: vb - Frequent PosterImage 3: polepole - Frequent PosterImage 4: _j - Most Recent PosterImage 5: windysoliloquy - Frequent PosterImage 6: Mauricio1997 - Most Recent Poster | 9 | 1.8k | 1…

• [15] Codex for (almost) everything | OpenAIopenai.com

  What’s next In just the year since Codex launched, the ways developers are using Codex has expanded. Developers start with Codex to write code, then increasingly use it to understand systems, gather context, review work, debug issues, coordinate with teammates, and keep longer-running work moving. Our mission is to ensure that AGI benefits all of humanity. That includes narrowing the gap between what people can imagine and what they can build. This release brings Codex closer to the tools, workflows, and decisions involved in building software, with much more to come soon. 2026 Codex ## Au…

• [16] GPT-5.3 and GPT-5.5 in ChatGPT | OpenAI Help Centerhelp.openai.com

  GPT-5.3 and GPT-5.5 in ChatGPT | OpenAI Help Center Image 1: OpenAI Language English United States Login 1. All Collections 2. ChatGPT 3. GPT-5.3 and GPT-5.5 in ChatGPT # GPT-5.3 and GPT-5.5 in ChatGPT Updated: 16 minutes ago As of February 13, 2026, models GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 (Instant and Thinking) have been retired from ChatGPT and are no longer available. API access remains unchanged. _ChatGPT Business, Enterprise, and Edu customers will retain access to GPT-4o within Custom GPTs until April 3, 2026. After April 3, GPT-4o will be fully retired acros…

• [17] GPT-5.5 Bio Bug Bounty | OpenAIopenai.com

  If you’re interested in supporting OpenAI’s work to deliver safe and secure artificial intelligence beyond the Bio Bounty program, you can learn about our Safety Bug Bounty⁠(opens in a new window) and Security Bug Bounty⁠(opens in a new window)⁠ programs. ## Keep reading View all Image 1: System Card Card SEO 1x1 GPT-5.5 System Card Safety Apr 23, 2026 Image 2: accelerating-cyber-defense-ecosystem-1x1 Accelerating the cyber defense ecosystem that protects us all Security Apr 16, 2026 Image 3: Scaling our trusted access program for cyber defense 1x1 Trusted access for the next era of cyber def…

• [18] GPT-5.5 is here! Available in Codex and ChatGPT todaycommunity.openai.com

  Announcements models You have selected 0 posts. select all cancel selecting 3.7k views 35 likes 2 links 8 users Image 2: polepole2 Image 3: Espresso Bean Image 4: alonso quintanilla Image 5: Mauricio Barros Image 6 Summarize Apr 23 1 / 10 Apr 24 5h ago ## post by vb 8 hours ago Image 7 vb Leader Image 8: potato 3 8h Introducing GPT-5.5 A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done. Image 9: HGm8jVWbsAAwL60 HGm8jVWbsAAwL60 1…

• [19] GPT-5.5 is here! Available in Codex and ChatGPT today - #9 by af0rcommunity.openai.com

  GPT-5.5 is here! Available in Codex and ChatGPT today AGI IS HERE BOYS LETSGO fast mode will go off tho this model is more expensive keep that in mind! ### Related topics | Topic | | Replies | Views | Activity | --- --- | Codex using up MASSIVE credits (850 credits + 5 hour limit used on only 8 queries) Codex codex , bugs | 18 | 5450 | February 26, 2026 | | Token Pricing Trends for average users Codex CLI | 2 | 319 | August 9, 2025 | | Introducing New $100/month Pro Tier Announcements | 21 | 5494 | April 15, 2026 | | Hitting Codex limits after a few prompts Codex codex | 8 | 993 | January 7…

• [20] GPT-5.5 System Cardopenai.com

  GPT-5.5 System Card | OpenAI Skip to main content Log inTry ChatGPT(opens in a new window) Research Products Business Developers Company Foundation(opens in a new window) GPT-5.5 System Card | OpenAI April 23, 2026 SafetyPublication # GPT‑5.5 System Card Read the System Card(opens in a new window) Share ## 1. Introduction GPT‑5.5 is a new model designed for complex, real-world work, including writing code, researching online, analyzing information, creating documents and spreadsheets, and moving across tools to get things done. Relative to earlier models, GPT‑5.5 understands the task earlie…

• [21] Making ChatGPT better for clinicians - OpenAIopenai.com

  Keep reading View all Image 2: Hero Art Card SEO 1x1 Introducing GPT-5.5 Product Apr 23, 2026 Image 3: OAI Blog Agents Hero 1x1 Introducing workspace agents in ChatGPT Product Apr 22, 2026 Image 4: Images 2.0 blog art card Introducing ChatGPT Images 2.0 Product Apr 21, 2026 Our Research Research Index Research Overview Research Residency Economic Research Latest Advancements GPT-5.5 GPT-5.4 GPT-5.3 Instant GPT-5.3-Codex Safety Safety Approach Security & Privacy Trust & Transparency ChatGPT Explore ChatGPT(opens in a new window) Business Enterprise Education Pricing(opens in a new window) D…

• [22] OpenAI Newsroom | Productopenai.com

  OpenAI Newsroom | Product | OpenAI Skip to main content Log inTry ChatGPT(opens in a new window) Research Products Business Developers Company Foundation(opens in a new window) Try ChatGPT(opens in a new window)Login OpenAI ## Product Company Research Product Safety Engineering Security Global Affairs AI Adoption All Filter Sort Switch cards to show Media Switch cards to hide Media Image 1: Hero Art Card SEO 1x1 Introducing GPT-5.5 Product Apr 23, 2026 Image 2: Making ChatGPT free for clinicians Making ChatGPT better for clinicians Product Apr 22, 2026 Image 3: OAI Blog Agents Hero 1x1 Intr…

• [23] GPT-5.5 is here! Available in Codex and ChatGPT todaycommunity.openai.com

  Related topics | Topic | | Replies | Views | Activity | --- --- | Introducing the New Codex for (almost) everything Announcements codex , codex-app | 24 | 2744 | April 23, 2026 | | GPT-5.1-Codex-Max is now available in the API Announcements | 11 | 2877 | December 11, 2025 | | Upgrades to Codex — gpt-5-codex Announcements codex , gpt-5-codex | 26 | 4017 | October 2, 2025 | | Introducing GPT-5.4 mini and nano — our most capable small models yet Announcements | 5 | 1769 | March 18, 2026 | | Introducing GPT-5.2-Codex Codex announcement , codex , chatgpt , api | 3 | 1190 | December 18, 2025 |…

• [24] Introducing GPT-5.5openai.com

  Introducing GPT-5.5 | OpenAI Skip to main content Log inTry ChatGPT(opens in a new window) Research Products Business Developers Company Foundation(opens in a new window) Try ChatGPT(opens in a new window)Login OpenAI Table of contents Model capabilities Next-generation inference efficiency Advancing cybersecurity for everyone’s safety Availability and pricing Evaluations April 23, 2026 ProductRelease # Introducing GPT‑5.5 A new class of intelligence for real work Loading… Share We’re releasing GPT‑5.5, our smartest and most intuitive to use model yet, and the next step toward a new way of…

• [25] Models | OpenAI APIdevelopers.openai.com

  Legacy APIs Assistants API Migration guide Deep dive Tools ### Resources Terms and policies Changelog Your data Permissions Rate limits Deprecations MCP for deep research Developer mode ChatGPT Actions Introduction Getting started Actions library Authentication Production Data retrieval Sending files # Models GPT-5.5 is currently available in ChatGPT and Codex, with API availability coming soon. ## Choosing a model If you're not sure where to start, use gpt-5.4, our flagship model for complex reasoning and coding. If you're optimizing for latency and cost, choose a smaller variant like gp…

• [26] API Pricing - OpenAIopenai.com

  OpenAI API Pricing | OpenAI Skip to main content Log inTry ChatGPT(opens in a new window) Research Products Business Developers Company Foundation(opens in a new window) OpenAI API Pricing | OpenAI # API Pricing Contact sales ## Flagship models Our frontier models are designed to spend more time thinking before producing a response, making them ideal for complex, multi-step problems. Choose your processing mode Standard Batch -50%Data residency +10% ## GPT-5.5 (coming soon) A new class of intelligence for coding and professional work. ### Price Input: $5.00 / 1M tokens Cached input: $0.50 /…

• [27] Introducing GPT-5 - OpenAIopenai.com

  Keep reading View all Image 1: Hero Art Card SEO 1x1 Introducing GPT-5.5 Product Apr 23, 2026 Image 2: Making ChatGPT free for clinicians Making ChatGPT better for clinicians Product Apr 22, 2026 Image 3: OAI Blog Agents Hero 1x1 Introducing workspace agents in ChatGPT Product Apr 22, 2026 Our Research Research Index Research Overview Research Residency Economic Research Latest Advancements GPT-5.5 GPT-5.4 GPT-5.3 Instant GPT-5.3-Codex Safety Safety Approach Security & Privacy Trust & Transparency ChatGPT Explore ChatGPT(opens in a new window) Business Enterprise Education Pricing(opens in…

• [28] Introducing GPT-5.4 | OpenAIopenai.com

  Introducing GPT-5.4 | OpenAI Skip to main content Log inTry ChatGPT(opens in a new window) Research Products Business Developers Company Foundation(opens in a new window) Try ChatGPT(opens in a new window)Login OpenAI Table of contents Knowledge work Computer use and vision Coding Tool use Steerability Safety Availability and pricing Evaluations March 5, 2026 ProductRelease # Introducing GPT‑5.4 Designed for professional work Loading… Share Today, we’re releasing GPT‑5.4 in ChatGPT (as GPT‑5.4 Thinking), the API, and Codex. It’s our most capable and efficient frontier model for professional…

• [29] Introducing workspace agents in ChatGPTopenai.com

  Making ChatGPT better for clinicians Product Apr 22, 2026 Image 6: Images 2.0 blog art card Introducing ChatGPT Images 2.0 Product Apr 21, 2026 Our Research Research Index Research Overview Research Residency Economic Research Latest Advancements GPT-5.5 GPT-5.4 GPT-5.3 Instant GPT-5.3-Codex Safety Safety Approach Security & Privacy Trust & Transparency ChatGPT Explore ChatGPT(opens in a new window) Business Enterprise Education Pricing(opens in a new window) Download(opens in a new window) Sora Sora Overview Features Pricing Sora log in(opens in a new window) [...] Teams do their best work w…

• [30] OpenAI Research | Releaseopenai.com

  OpenAI Research | Release | OpenAI Skip to main content Log inTry ChatGPT(opens in a new window) Research Products Business Developers Company Foundation(opens in a new window) Try ChatGPT(opens in a new window)Login OpenAI ## Research All Publication Conclusion Milestone Release Filter Sort Switch cards to show Media Switch cards to hide Media Product Apr 23, 2026 Introducing GPT-5.5 Introducing GPT-5.5, our smartest model yet—faster, more capable, and built for complex tasks like coding, research, and data analysis across tools. Research Apr 22, 2026 [...] Product Mar 5, 2026 Introducing…

• [31] The next phase of enterprise AI | OpenAIopenai.com

  Image 3: Frame OpenAI acquires TBPN Company Apr 2, 2026 Our Research Research Index Research Overview Research Residency Economic Research Latest Advancements GPT-5.5 GPT-5.4 GPT-5.3 Instant GPT-5.3-Codex Safety Safety Approach Security & Privacy Trust & Transparency ChatGPT Explore ChatGPT(opens in a new window) Business Enterprise Education Pricing(opens in a new window) Download(opens in a new window) Sora Sora Overview Features Pricing Sora log in(opens in a new window) API Platform Platform Overview Pricing API log in(opens in a new window) Documentation(opens in a new window) Developer…

• [32] OpenAI Newsopenai.com

  OpenAI News | OpenAI Skip to main content Log inTry ChatGPT(opens in a new window) Research Products Business Developers Company Foundation(opens in a new window) OpenAI News | OpenAI ## All Company Research Product Safety Engineering Security Global Affairs AI Adoption All Filter Sort Switch cards to show Media Switch cards to hide Media Image 1: Hero Art Card SEO 1x1 Introducing GPT-5.5 Product Apr 23, 2026 Image 2: System Card Card SEO 1x1 GPT-5.5 System Card Safety Apr 23, 2026 Image 3: GPT-5.5 Bio Bug Bounty > art card GPT-5.5 Bio Bug Bounty Safety Apr 23, 2026 Image 4: Making ChatGPT…

• [33] Anthropic releases Claude Opus 4.7, narrowly retaking lead for most ...venturebeat.com

  Knowledge Work (GDPVal-AA): It achieved an Elo score of 1753, notably outperforming GPT-5.4 (1674) and Gemini 3.1 Pro (1314). Agentic Coding (SWE-bench Pro): The model resolved 64.3% of tasks, compared to 53.4% for its predecessor. Graduate-Level Reasoning (GPQA Diamond): It reached 94.2%, maintaining parity with the industry's most advanced models while improving on its internal consistency. Visual Reasoning (arXiv Reasoning): With tools, the model scored 91.0%, a meaningful jump from the 84.7% seen in Opus 4.6. [...] # Anthropic releases Claude Opus 4.7, narrowly retaking lead for most powe…

• [34] Claude Opus 4.7 Benchmarks Explainedvellum.ai

  Apr 16, 2026•16 min•ByNicolas Zeeb Guides CONTENTS Key observations of reported benchmarks Coding capabilities SWE-bench Verified SWE-bench Pro Terminal-Bench 2.0 Agentic capabilities MCP-Atlas (Scaled tool use) Finance Agent v1.1 OSWorld-Verified (Computer use) BrowseComp (Agentic search) Reasoning capabilities GPQA Diamond (Graduate-level science) Humanity's Last Exam Multimodal and vision capabilities CharXiv Reasoning (Visual reasoning) Multilingual Q&A (MMMLU) Safety and alignment What these benchmarks really mean for your agents When to use Opus 4.6 vs Opus 4.7 Use Opus 4.7 with your Ve…

• [35] Claude Opus 4.7: Model Specifications and Detailsapxml.com

  ApX logoApX logo # Claude Opus 4.7 Models Claude 4.7 → Claude Opus 4.7 Context Length 200K Modality Multimodal Architecture Dense License Proprietary Release Date 16 Apr 2026 Knowledge Cutoff - ### Technical Specifications Attention Structure Multi-Head Attention Hidden Dimension Size - Number of Layers - Attention Heads - Key-Value Heads - Activation Function - Normalization - Position Embedding Absolute Position Embedding ### Claude Opus 4.7 [...] Absolute Position Embedding ### Claude Opus 4.7 Anthropic's most capable Claude 4.7 model, offering a notable improvement on Opus 4.6 in advanced…

• [36] Claude Opus 4.7: What Changed for Coding Agents (April 2026)verdent.ai

  | Benchmark | Opus 4.6 | Opus 4.7 | GPT-5.4 | Notes | --- --- | SWE-bench Verified | 80.80% | 87.60% | — | Anthropic-conducted; memorization screens applied | | SWE-bench Pro | 53.50% | 64.30% | 57.70% | Multi-language real-world tasks | | CursorBench | 58% | 70% | — | Source: Cursor CEO Michael Truell (partner eval) | | Terminal-Bench 2.0 | — | 69.40% | 75.10% | GPT-5.4 leads; Terminus-2 harness, thinking disabled | | BrowseComp | 83.70% | 79.30% | 89.30% | Regression vs Opus 4.6 | | XBOW Visual Acuity | 54.50% | 98.50% | — | Computer use / screenshot tasks | | GPQA Diamond | 91.30% | 94.20%…

• [37] GPT-5.5 vs Claude Opus 4.7: Pricing, Speed, Benchmarks - LLM Statsllm-stats.com

  | Spec | GPT-5.5 | Claude Opus 4.7 | --- | Provider | OpenAI | Anthropic | | Release date | Apr 23, 2026 | Apr 16, 2026 | | Model ID | gpt-5.5 | claude-opus-4-7 | | Input / output (≤200K) | $5 / $30 per 1M | $5 / $25 per 1M | | Input / output (>200K) | $5 / $30 per 1M (flat) | $10 / $37.50 per 1M | | Context window (input / output) | 1M / 128K | 1M / 128K | | Modalities | Text + image, text out | Text + image (~3.75 MP), text out | | Reasoning controls | xhigh effort tier | low / medium / high / xhigh / max | | Batch / Flex tier | 0.5× standard | 0.5× standard | | Self-verification on age…

• [38] GPT‑5.5 vs. Claude Opus 4.7: A Benchmark-by-Benchmark Field Guide to the New Frontier - Kingy AIkingy.ai

  Curtis Pyke by Curtis Pyke When Anthropic shipped Claude Opus 4.7 on April 16, 2026 and OpenAI responded one week later with GPT‑5.5 on April 23, 2026, the frontier-model leaderboard shuffled twice in seven days. Both vendors are pitching these systems as flagships for coding, agentic work, scientific research, and professional knowledge tasks. But the moment you stop reading the marketing and start reading the benchmark tables, the story gets messy: GPT‑5.5 wins more head-to-heads than Opus 4.7 in official launch materials, while Opus 4.7 still wins on several of the benchmarks enterprise bu…

• [39] OpenAI's GPT-5.5 masters agentic coding with 82.7% benchmark ...interestingengineering.com

  On SWE-Bench Pro, it reached 58.6%, solving more real-world GitHub issues in a single pass than earlier versions. The model also outperformed its predecessor in long-horizon engineering tasks measured by internal benchmarks. These tasks often take human developers up to 20 hours to complete. > Introducing GPT-5.5 > > A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done. > > Now available in ChatGPT and Codex. pic.twitter.com/rPLTk…

• [40] Claude Opus 4.7 Is Here — What Changed, What's Better, and Is It Worth Upgrading?miraflow.ai

  The competitive landscape in April 2026 is intense. Here is how the three major frontier models stack up. opus-performance-chart.png GPT-5.4 trades blows with Opus 4.7 depending on the task, and Gemini 3.1 Pro holds its own on multilingual benchmarks. But on the aggregate, particularly for agentic and coding workloads where Claude has historically led, Opus 4.7 extends the gap rather than ceding ground.( On SWE-bench Pro (agentic coding), Opus 4.7 leads at 64.3% compared to GPT-5.4 at 57.7% and Gemini 3.1 Pro at 54.2%. On graduate-level reasoning (GPQA Diamond), all three models score in the…

• [41] GPT-5.5: Pricing, Benchmarks & Performancellm-stats.com

  9Image 42GPT-5 mini 0.22 10Image 43o3 0.16 GPQAView → #4 of 10 Image 44: LLM Stats Logo A challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. Questions are Google-proof and extremely difficult, with PhD experts reaching 65% accuracy. More 1Image 45Claude Mythos Preview 0.95 2Image 46Gemini 3.1 Pro 0.94 3Image 47Claude Opus 4.7 0.94 4Image 48GPT-5.5 0.94 5Image 49GPT-5.2 Pro 0.93 6Image 50GPT-5.4 0.93 7Image 51GPT-5.2 0.92 8Image 52Gemini 3 Pro 0.92 9Image 53Claude Opus 4.6 0.91 10Image 54Kimi K2.6 0.91 Show 18 more Notice missing…

• [42] OpenAI’s GPT-5.5 benchmarks show a 60% hallucination drop and coding skills that rival senior engineers – Startup Fortunestartupfortune.com

  The numbers are hard to argue with. GPT-5.5 scored 92.4% on the MMLU benchmark, up from GPT-4’s 86.4%, and hit 88.7% on SWE-bench, the industry’s most demanding coding evaluation. That SWE-bench figure effectively places the model at senior software engineer level for resolving real GitHub issues, not toy problems. OpenAI announced the results via a joint livestream on X and its official blog, confirming that enterprise API subscribers get access immediately, with consumer rollout scheduled for early May.

• [43] Anthropic releases Claude Opus 4.7: How to try it, benchmarks, safetymashable.com

  Anthropic releases Claude Opus 4.7: How to try it, benchmarks, safety headshot of timothy beck werth, a handsome journalist with great hair The Claude AI logo is displayed on a smartphone screen with a multitude of Anthropic logos in the background Anthropic has been shipping products and making news at a blistering pace in 2026, and on Thursday, the AI company announced the launch of Claude Opus 4.7. Claude Opus 4.7 is Anthropic's most intelligent model available to the general public. Notably, Anthropic said in a press release") that Opus 4.7 is not as powerful as Claude Mythos, which Ant…

• [44] Claude Opus 4.7 is Out — Weekly AI Newsletter (April 20th 2026)medium.com

  TRACER: Trace-Based Adaptive Cost-Efficient Routing for LLM Classification. TRACER uses lightweight execution traces to route classification requests across model tiers, cutting cost while preserving accuracy. The method adapts routing decisions to observed difficulty patterns rather than static rules, outperforming fixed-threshold baselines in deployment-scale LLM classification workloads. [...] Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation. The survey consolidates research on the attention sink phenomenon, where transformers route disproportionate a…