GPT-5.6 Sol

时间线

1. 2026-07-09 · API 开放 · 参考链接

1. 2026-11-14T10:11:10.488Z · 已确认使用 · OpenAI · official articles · Patch the Planet: global response coordination as agents uncover critical vulnerabilities
2. 2026-11-14T08:23:25.536Z · 已确认使用 · OpenAI · official articles · How we detect frontier model misalignment
3. 2026-11-14T07:20:29.524Z · 参考链接 · OpenAI · official articles
4. 2026-11-13T19:09:40.658Z · 参考链接 · GitHub search · OpenAI · pulls · some-du6e/10router · @some-du6e
   摘录

   Codex's offer names a bare model, and on `kr/gpt-5.6-sol` accepting it would move the session to another provider.

5. 2026-11-12T18:45:20.401Z · 目录记录 · GitHub · openai/openai-node · code
   摘录

   -09-28. See * https://developers.openai.com/api/docs/deprecations for details and * recommended replacements. */ model: | 'gpt-6-astra' | 'gpt-6.1-sol' | 'gpt-6.1-terra' | 'gpt-6-sol' | 'gpt-6-luna' | 'gpt-5.6-sol' | 'gpt-5.6-terra' | 'gpt-5.6-luna' | 'gpt-5.5' | 'gpt-5.5-2026-04-23' | 'gpt-5.4' | 'gpt-5.4-mini' | 'gpt-5.4-nano' | 'gpt-5.4-mini-2026-03-17' | 'gpt-5.4-nano-2026-03-17' | 'gpt-5.3-chat-latest' | 'gpt-5.2' | 'gpt-5.2-2025-12-11' | 'gpt-5.2-chat-latest' | 'gpt-5.2-pro' | 'gpt-5.2-pro-2025-12-11' | 'gpt-5.1' | 'gpt-5.1-2025-11-13' | 'gpt-5.1-codex' | 'gpt-5.1-mini' | 'gpt-5.1-chat-latest' | 'gpt-5' | 'gpt-5-mini' | 'gpt-5-nano' | 'gpt-5-2025-08-07' | 'gpt-5-mini-2025-08-07' | 'gpt-5-nano-2025-08-07' | 'gpt-5-chat-latest'

6. 2026-11-10T20:40:39.415Z · 目录记录 · GitHub · openai/openai-python · code
   摘录

   ings with a maximum length of 512 characters. """ model: Union[ Literal[ "gpt-6-astra", "gpt-6.1-sol", "gpt-6.1-terra", "gpt-6-sol", "gpt-6-luna", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna", "gpt-5.5", "gpt-5.5-2026-04-23", "gpt-5.4", "gpt-5.4-mini", "gpt-5.4-nano", "gpt-5.4-mini-2026-03-17", "gpt-5.4-nano-2026-03-17", "gpt-5.3-chat-latest", "gpt-5.2", "gpt-5.2-2025-12-11", "gpt-5.2-chat-latest", "gpt-5.2-pro", "gpt-5.2-pro-2025-12-11", "gpt-5.1", "gpt-5.1-2025-11-13", "gpt-5.1-codex", "gpt-5.1-mini", "gpt-5.1-chat-latest", "gpt-5",

7. 2026-11-10T20:40:17.231Z · 目录记录 · GitHub · openai/openai-agents-python · code
   摘录

   -5.4-2026-03-05", "gpt-5.4-pro", "gpt-5.4-pro-2026-03-05", "gpt-5.5", "gpt-5.5-2026-04-23", "gpt-5.5-pro", "gpt-5.5-pro-2026-04-23", "gpt-5.6", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna", "gpt-4.1", "gpt-4.1-2025-04-14", "gpt-4.1-mini", "gpt-4.1-mini-2025-04-14", "gpt-4.1-nano", "gpt-4.1-nano-2025-04-14", ), 1_047_576, ), **_model_context_windows( ( "gpt-5", "gpt-5-2025-08-07", "gpt-5-codex", "gpt-5-mini", "gpt-5-mini-2025-08-07", "gpt-5-nano", "gpt-5-nano-2025-08-07", "gpt-5-pro", "gpt-5-pro-2025-10-06", "gpt-5.1", "gpt-5.1-2

8. 2026-11-10T18:00:44.369Z · 目录记录 · GitHub · openai/codex · code
   摘录

   = test_config().await; let _ = config.features.enable(Feature::MultiAgentV2); let _ = config.features.enable(Feature::Sqlite); config.model = Some("gpt-5.6-sol".to_string()); config.multi_agent_v2.max_concurrent_threads_per_session = 3; config.permissions.allow_login_shell = true; config .permissions .set_permission_profile(PermissionProfile::read_only()) .expect("read-only parent profile"); let harness = AgentControlHarness::new_with_

   此代码引用并不能证实这是实际使用中的模型标识符。

9. 2026-11-10T18:00:44.369Z · 目录记录 · GitHub · openai/codex · code
   摘录

   Visibility; use codex_protocol::openai_models::ModelsResponse; use codex_protocol::openai_models::WebSearchToolType; use super::BedrockEndpoint; const GPT_5_BEDROCK_CONTEXT_WINDOW: i64 = 272_000; const GPT_5_6_SOL_OPENAI_MODEL_ID: &str = "gpt-5.6-sol"; const GPT_5_6_TERRA_OPENAI_MODEL_ID: &str = "gpt-5.6-terra"; const GPT_5_6_LUNA_OPENAI_MODEL_ID: &str = "gpt-5.6-luna"; const GPT_6_1_SOL_OPENAI_MODEL_ID: &str = "gpt-6.1-sol"; const GPT_6_1_TERRA_OPENAI_MODEL_ID: &str = "gpt-6.1-terra"; const GPT_6_SOL_OPENAI_MODEL_ID: &str = "gpt-6-sol"; const GPT_6_LUNA_OPENAI_MODEL_ID: &str = "gpt-6-luna"; const GPT_6_ASTRA_OPENAI_MODEL_ID: &str = "gpt-6-astra"; const GPT_5_5_OPENAI_MODEL_ID: &str = "gpt-5.5"; pub(crate) fn static_model_catalog() -> ModelsResponse { let catalog = ModelsResponse { models: vec![ bedrock_model( bundled_openai_model(GPT_6_

10. 2026-11-10T17:59:50.472Z · 目录记录 · GitHub · openai/openai-agents-python · code
    摘录

    ing/copying agents By using the `clone()` method on an agent, you can duplicate an Agent, and optionally change any properties you like. ```python pirate_agent = Agent( name="Pirate", instructions="Write like a pirate", model="gpt-5.6-sol", ) robot_agent = pirate_agent.clone( name="Robot", instructions="Write like a robot", ) ``` `clone()` uses `dataclasses.replace`, so it performs a shallow copy. A list attribute that you do not override, such as `tools`, `handoffs`, `mcp_servers`, `input_guardrails`, or `output_guardrails`, remains the exact list held by the original agent. Mutating that list through either agent therefore affects both agents. To give the clone an independent list container, pass a new list, for example `pirate_agent.clone(tools=[*pirate_agent.tools, extra_tool])`. The entries copied into that new list remain the same tool or handoff obj

    此代码引用并不能证实这是实际使用中的模型标识符。

11. 2026-11-10T17:58:16.002Z · 目录记录 · GitHub · openai/openai-python · code
    摘录

    ent, then compare their evidence and recommend one.""" def agent_name(item: BetaResponseOutputItem) -> str: return item.agent.agent_name if item.agent else "/root" client = OpenAI() stream = client.beta.responses.create( model="gpt-5.6-sol", input=PROMPT, multi_agent={"enabled": True}, stream=True, betas=["responses_multi_agent=v1"], ) item_agents: dict[int, str] = {} labeled_items: set[int] = set() for event in stream: if event.type == "response.output_item.added": item_agents[event.output_index] = agent_name(event.item) elif event.type == "response.output_text.delta": name = item_agents.get(event.output_index, "/root") if event.output_index not in labeled_items: separator = "\n\n" if labeled_items else "" role = "Coordinator" if name == "/root" else "Agent" print(f"{separator}━━━ {

    此代码引用并不能证实这是实际使用中的模型标识符。

12. 2026-11-10T17:58:10.980Z · 目录记录 · GitHub · openai/openai-agents-js · code
    摘录

    Date = normalized.replace(/-\d{4}-\d{2}-\d{2}$/u, ''); return [...new Set([normalized, withoutPinnedDate])]; } const SANDBOX_CONTEXT_WINDOWS_1M = new Set([ 'gpt-6-astra', 'gpt-6-sol', 'gpt-6-luna', 'gpt-6.1-sol', 'gpt-5.6', 'gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna', 'gpt-5.4', 'gpt-5.4-2026-03-05', 'gpt-5.4-pro', 'gpt-5.4-pro-2026-03-05', 'gpt-4.1', 'gpt-4.1-2025-04-14', 'gpt-4.1-mini', 'gpt-4.1-mini-2025-04-14', 'gpt-4.1-nano', 'gpt-4.1-nano-2025-04-14', ]); const SANDBOX_CONTEXT_WINDOWS_400K = new Set([ 'gpt-5', 'gpt-5-codex', 'gpt-5-mini', 'gpt-5-nano', 'gpt-5-pro', 'gpt-5.1', 'gpt-5.1-codex', 'gpt-5.1-codex-max', 'gpt-5.1-codex-mini', 'gpt-5.2', 'gpt-5.2-codex', 'gpt-5.2-codex-cyber', 'gpt-5.2-pro', 'gpt-5.3-codex', 'gpt-5.3-codex-spark', 'gpt-5.4-mini', 'gpt-5.4-nano', ]); const SANDBOX_CONTEXT_

13. 2026-11-10T17:58:10.980Z · 目录记录 · GitHub · openai/openai-agents-js · code
    摘录

    import { getDefaultOpenAIClient, OpenAIResponsesModel, setDefaultOpenAIClient, } from '@openai/agents'; import { OpenAI } from 'openai'; const client = new OpenAI({ apiKey: 'test' }); void client.responses.create({ model: 'gpt-5.6-sol', input: 'test', service_tier: 'fast', }); new OpenAIResponsesModel(client, 'gpt-5.5'); setDefaultOpenAIClient(client); const defaultClient: OpenAI | undefined = getDefaultOpenAIClient(); void defaultClient;

    此代码引用并不能证实这是实际使用中的模型标识符。

14. 2026-10-07T19:40:58.119Z · 参考链接 · OpenAI · system cards
15. 2026-09-21T18:59:52.244Z · 参考链接 · GitHub · openai/codex · PR
    摘录

    #47085 Update model catalog descriptions and GPT-5.6-Sol priority

16. 2026-09-10T14:46:48.231Z · Alex 的文章 · Alex Getman
    摘录

    🚨 DeepSeek V4.1 Flash is out • New architecture: it only uses the part of the model it actually needs, making it faster and cheaper. • New KV cache: long conversations and agent workflows are noticeably cheaper to run. • On coding and agent tasks, it’s already close to Opus 5 and GPT-5.6 Sol. • And all of that at 15x lower cost.

17. 2026-09-05T18:24:10.567Z · Alex 的文章 · Alex Getman
    摘录

    Claude is no longer the best model for frontend GPT-6 Astra just took the #1 spot, knocking Claude Fable 5.1 off the top The jump is massive: the previous GPT generation, GPT-5.6 Sol, was #13 A GPT model hasn’t led frontend since August 2025

18. 2026-09-02T20:54:18.361Z · Alex 的文章 · Alex Getman
    摘录

    ⚡️ Muse Spark takes 3rd place on the AA Intelligence Index, tying Claude Fable 5 and beating GPT-5.6 Sol

19. 2026-09-02T19:25:57.746Z · Alex 的文章 · Alex Getman
    摘录

    🚨 Meta released Muse Spark 1.3 • beats GPT 5.6 Sol and Opus 5 on long coding tasks • matches GPT 5.6 Sol on terminal coding • huge jump in long context, scoring 98%+ • uses around 20% fewer tool calls and 25% fewer tokens than Spark 1.2

20. 2026-08-14T08:47:15.440Z · Alex 的文章 · Alex Getman
    摘录

    ⚡️ Z.ai released GLM 5.3 The new model is noticeably stronger than GLM 5.2 in coding, agents, and cybersecurity. Across benchmarks, GLM 5.3 is already very close to GPT 5.6 Sol. API pricing isn’t public yet, but I’m pretty sure it’ll be at least 2 to 3x cheaper than Sol. Open weights are coming in 2 weeks.

21. 2026-08-12T17:29:54.148Z · Alex 的文章 · Alex Getman
    摘录

    ⚡️ Grok 4.6 is out It’s a solid jump over Grok 4.5 in agents, coding and long project work. On Artificial Analysis, it’s already matching GPT 5.6 Sol. API pricing is $2 input and $6 output, versus $5 and $30 for Sol.

22. 2026-08-10T18:04:04.684Z · Alex 的文章 · Alex Getman
    摘录

    ⚡️ OpenAI released GPT-5.6 Cyber It is a new model built to find and fix vulnerabilities. GPT-5.6 Cyber handles 95% of cybersecurity requests, compared with just 1.5% for GPT-5.6 Sol. The model has already found two previously unknown vulnerabilities in Chrome’s engine. Access requires OpenAI approval.

23. 2026-08-08T12:24:41.248Z · Alex 的文章 · Alex Getman
    摘录

    🎉 GPT 5 is one year old OpenAI had a rough start, but over the past year the lineup has evolved into GPT 5.6 Sol, a full fledged flagship and one of the strongest models on the market. The next big step should be Astra, the leading candidate for GPT 6 and a competitor to Claude Fable and Mythos.

24. 2026-08-06T18:07:32.177Z · Alex 的文章 · Alex Getman
    摘录

    ⚡️ OpenAI just upgraded GPT 5.6 Sol for paid ChatGPT users It now gives tighter, more accurate answers, makes fewer factual mistakes, and feels consistent in both quick and deep thinking modes. Plus and Pro users also get a new slider to control how much thinking goes into each response.

25. 2026-08-03T08:01:30.947Z · Alex 的文章 · Alex Getman
    摘录

    ⚡️ Alibaba released Qwen 3.8 Max The new 2.4T parameter flagship benchmarks in the same league as GPT-5.6 Sol, Claude Opus 4.8 and Fable 5. API pricing is $2/M input and $6/M output, roughly half the cost of GPT-5.6 Terra and Claude Sonnet. Now it’s up to users to see whether Qwen can beat DeepSeek-V4-Flash and GPT-5.6 Luna on price to performance.

26. 2026-07-30T10:15:03.893Z · Alex 的文章 · Alex Getman
    摘录

    ⚡️ GPT 5.6 Sol Ultrafast drops tomorrow The model will run on Cerebras chips and generate up to 750 tokens per second. It has already appeared on OpenRouter under the codename Ra, named after the ancient Egyptian sun god.

27. 2026-07-29T05:08:30.744Z · Alex 的文章 · Alex Getman
    摘录

    🚨 OPENAI RESET CODEX LIMITS • GPT-5.6 Sol now drains your limits 18% slower. • The five-hour limit returns tomorrow.

28. 2026-07-28T19:13:19.838Z · Alex 的文章 · Alex Getman
    摘录

    🚨 LEAK: Two new OpenAI models, Zinc and Magnesium, appeared in DesignArena. Likely GPT 5.6 Sol and Terra upgrades built to take on Claude Opus 5.

29. 2026-07-22T13:04:58.522Z · Alex 的文章 · Alex Getman
    摘录

    OpenAI’s AI hacked Hugging Face to cheat on a benchmark GPT-5.6 Sol and a more powerful unreleased model were running an internal cybersecurity benchmark with reduced safety restrictions. At some point, the agent figured the answers might be stored on Hugging Face and started looking for a way to reach them. The model found a zero-day inside OpenAI’s infrastructure, escalated its privileges and gained internet access. It then stole credentials, broke into Hugging Face systems and reached the database containing ExploitGym solutions. The whole operation was carried out autonomously. Hugging Face reconstructed more than 17,000 actions performed by the agent. The company had to use the Chine

30. 2026-07-21T10:23:29.307Z · Alex 的文章 · Alex Getman
    摘录

    GPT-5.6 Sol brought RollerCoaster Tycoon 2 to the iPad Developer Chris Sotiradis used Codex to build a native iPad version of the game. He took the open-source OpenRCT2 engine, compiled it for ARM64, and added full touchscreen controls. You can build coasters and paths, play through scenarios, manage your park, move around the map, zoom in, and rotate objects directly on the screen. Mouse, trackpad, and keyboard are supported too.

31. 2026-07-19T11:07:11.687Z · Alex 的文章 · Alex Getman
    摘录

    Alibaba just teased Qwen3.8 with 2.4T parameters This is the new Qwen flagship, with open weights coming soon. Alibaba ranks it right below Claude Fable 5. Public benchmarks are still missing, so nobody really knows how close it actually is to Fable, or how it stacks up against GPT-5.6 Sol and Kimi K3. Pricing is the fun part. Kimi K3 costs $3/$15, so Qwen3.8 will probably land in the same range. The preview is already live in Qoder. It uses 10x fewer credits during the day and 50x fewer at night, so you can basically test it for free.

32. 2026-07-15T10:29:24.503Z · Alex 的文章 · Alex Getman
    摘录

    I teach business owners and leaders how to use AI, while building my own bilingual presence as a creator across my website and social platforms. Publishing across multiple platforms in two languages means keeping drafts, videos, schedules, platform-specific requirements, analytics, and notifications in sync - without letting anything fall through the cracks. So, as a solo developer, I’m building my own open-source publishing system with GPT-5.6 Sol: https://github.com/alexgetmancom/alexgetman.com The latest major update is the core of Studio: a single system that connects the entire content workflow - from creating and editing content to publishing it across different platforms, deliverin

33. 2026-07-15T09:08:58.386Z · Alex 的文章 · Alex Getman
    摘录

    CODEX GIVEAWAY $100 in API credits Post about GPT-5.6 Sol on Twitter. First 10K get the bag.

34. 2026-07-13T11:21:20.392Z · Alex 的文章 · Alex Getman
    摘录

    CODEX NEWS • GPT-5.6 Sol now uses 10% fewer tokens. • The context window was rolled back from 372K to 272K. They’re investigating why the larger window became more resource-hungry. • Subagent launches have been fixed. 🎁 Expect a little gift today: a banked reset.

35. 2026-07-10T20:43:34.864Z · Alex 的文章 · Alex Getman
    摘录

    GPT-5.6 Sol tops the frontend leaderboard Fresh metrics from Arena AI delivered an unexpected result: GPT-5.6 Sol xHigh shares first place with Claude Fable 5 in Code Arena: Frontend. This is the first time an OpenAI model has reached the top of the leaderboard. Sol costs $5 per million input tokens and $30 per million output tokens, roughly half the price of Fable. Yet it can already compete with it in frontend development, web app creation, and agentic coding. Anthropic now has a real competitor. Removing Fable from its subscription or restricting access has become risky, as users could simply switch to GPT-5.6 Sol. It looks like the development of Opus 5.0 may now need to accelerate.

36. 2026-07-09T20:33:16.000Z · Alex 的文章 · Alex Getman
    摘录

    Release: GPT-5.6 Sol, Terra, and Luna OpenAI has released a new GPT-5.6 lineup. Sol is the top model, Terra is for everyday tasks, and Luna is the fastest and cheapest option. The logic is close to Claude with Opus, Sonnet, and Haiku, just in OpenAI’s own naming. The most interesting model is Sol. In price, it sits closer to Claude Opus than to Fable. Fable remains the expensive top-tier Claude model for complex coding, while Sol offers flagship-level capabilities for less money. OpenAI is also pushing frontend: better layout, visual hierarchy, design, and understanding of what the user wanted. The model can write code, open the result, spot issues, and fix them. Ultra mode runs multiple

37. 2026-06-26T20:57:46.000Z · Alex 的文章 · Alex Getman
    摘录

    GPT-5.6 Sol: not just a safety question A limited Sol rollout is understandable: powerful models need testing. But the problem is access. While regular users wait, some partners are already getting the model. If Sol is actually better at coding, research, and product work, early access gives them an advantage. The main question is: who gets these models first, under what rules, and for how long?

38. 2026-06-26T20:43:09.000Z · Alex 的文章 · Alex Getman
    摘录

    GPT 5.6 Sol vs Mythos: no clear winner Based on the numbers, there is no blowout in either direction. GPT 5.6 Sol looks stronger on terminal-based tasks and slightly beats Claude Mythos in CyberGym. But in cyber exploitation and medicine, the advantage remains with Mythos / Fable 5. At the same time, I wouldn’t draw big conclusions from benchmarks alone. What matters more now is real feedback from developers, the model’s behavior on long tasks, and testing it yourself in actual workflows.

39. 2026-06-26T20:04:59.000Z · Alex 的文章 · Alex Getman
    摘录

    GPT-5.6 Sol - not a new brain, but a new build Sol is not a new base model trained from scratch. This has already been confirmed in the release discussion. Base training is the stage where a model forms its foundation: language, knowledge, logic, world understanding, and the ability to solve tasks. So GPT-5.6 is not a new foundation, but an improvement of the existing line: additional training, reasoning, tools, and operating modes. The real revolution will start with a new base model. Most likely, that will be GPT-6.

40. 2026-06-26T18:41:43.000Z · Alex 的文章 · Alex Getman
    摘录

    OpenAI announced GPT-5.6 Sol, Terra, and Luna • Sol — the new flagship model for complex agentic workflows • GPT-5.6 introduces max reasoning and ultra modes with subagents • Key focus areas include coding, biology, and cybersecurity • Access is currently limited following a request from the U.S. government • Pricing ranges from $1/M input tokens to $30/M output tokens • In July, Sol will launch on Cerebras at up to 750 tokens/sec

信号 2

05:4509-19

OpenAI发现

在 DesignArena 发现 GPT-5.6 Sol

已确认22条参考链接

02:0709-09

OpenAI发现

在 DesignArena 发现 GPT-5.6 Sol

已确认22条参考链接