Alex Getmanearly AI signals
EN

GPT-5.6 Sol

OpenAI

Released Updated
Context window
1M
1,050,000 tokens
Max output
128K
128,000 tokens
Input / Output
text, image → text

API pricing

USD / 1M tokens

InputOutputCached input
  • $4.00
openrouter
  • $20.00
openrouter
  • $0.40
openrouter
Calculate your workload

Comparing GPT-5.6 Sol on repository work

Sources checked:

OpenAI documents both gpt-5.6-sol and the gpt-5.6 alias that routes to it. Save the identifier actually used by your provider when comparing results. The catalogue identity on this page and an endpoint's API string serve different purposes; cloud providers can use their own identifiers.

For an agent trial, start with one bug and a reproducible failure. Keep the starting repository, instructions, tools and acceptance check the same for every run. Record the reasoning setting, failed attempts, time spent repairing the result and the final accepted change. The Codex usage worksheet helps measure work within one allowance window; the tool comparison adds subscription and review costs.

Check the billing conditions again before paying. The official reference qualifies long-context requests and promotional pricing. Use this page's provider evidence and the official price list for your exact route. An API rate does not establish how many tasks a ChatGPT or Codex subscription will complete.

Timeline

  1. · API availability · evidence link
Input / output price · USD / 1M tokens
ModelEvidence linksInput / output price
gpt-5.6-sol
batch · long context
openai
API route unavailable
$4.00 (Input > 272,000 tokens) / $15.00 (Input > 272,000 tokens)
gpt-5.6-sol
batch · short context
openai
API route unavailable
$2.00 (Input ≤ 272,000 tokens) / $10.00 (Input ≤ 272,000 tokens)
gpt-5.6-sol
fast · long context
openai
API route unavailable
$16.00 (Input > 272,000 tokens) / $60.00 (Input > 272,000 tokens)
gpt-5.6-sol
fast · short context
openai
API route unavailable
$8.00 (Input ≤ 272,000 tokens) / $40.00 (Input ≤ 272,000 tokens)
gpt-5.6-sol
flex · long context
openai
API route unavailable
$4.00 (Input > 272,000 tokens) / $15.00 (Input > 272,000 tokens)
gpt-5.6-sol
flex · short context
openai
API route unavailable
$2.00 (Input ≤ 272,000 tokens) / $10.00 (Input ≤ 272,000 tokens)
gpt-5.6-sol
standard · long context
openai
API route unavailable
$8.00 (Input > 272,000 tokens) / $30.00 (Input > 272,000 tokens)
gpt-5.6-sol
standard · short context
openai
API route unavailable
$4.00 (Input ≤ 272,000 tokens) / $20.00 (Input ≤ 272,000 tokens)
openai/gpt-5.6-sol
standard
openrouter$4.00 / $20.00
openai/gpt-5.6-sol-pro
standard
openrouter$4.00 / $20.00
openai/gpt-5.6-sol-pro:batch
batch
openrouter$2.00 / $10.00
openai/gpt-5.6-sol:batch
batch
openrouter$2.00 / $10.00
openai/gpt-5.6-sol
standard
vercel-gateway$4.00 / $20.00
openai/gpt-5.6-sol-batch
batch
vercel-gateway$2.00 / $10.00
  1. · Confirmed use · OpenAI · official articles · Patch the Planet: global response coordination as agents uncover critical vulnerabilities
  2. · Confirmed use · OpenAI · official articles · How we detect frontier model misalignment
  3. · Evidence links · OpenAI · official articles
  4. · Evidence links · GitHub search · OpenAI · pulls · some-du6e/10router · @some-du6e
    Excerpt
    Codex's offer names a bare model, and on `kr/gpt-5.6-sol` accepting it would move the session to another provider.
  5. · Catalogue listing · GitHub · openai/openai-node · code
    Excerpt
    -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. · Catalogue listing · GitHub · openai/openai-python · code
    Excerpt
    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. · Catalogue listing · GitHub · openai/openai-agents-python · code
    Excerpt
    -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. · Catalogue listing · GitHub · openai/codex · code
    Excerpt
    = 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_

    This code reference does not establish a working model identifier.

  9. · Catalogue listing · GitHub · openai/codex · code
    Excerpt
    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. · Catalogue listing · GitHub · openai/openai-agents-python · code
    Excerpt
    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

    This code reference does not establish a working model identifier.

  11. · Catalogue listing · GitHub · openai/openai-python · code
    Excerpt
    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}━━━ {

    This code reference does not establish a working model identifier.

  12. · Catalogue listing · GitHub · openai/openai-agents-js · code
    Excerpt
    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. · Catalogue listing · GitHub · openai/openai-agents-js · code
    Excerpt
    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;

    This code reference does not establish a working model identifier.

  14. · Evidence links · OpenAI · system cards
  15. · Evidence links · GitHub · openai/codex · PR
    Excerpt
    #47085 Update model catalog descriptions and GPT-5.6-Sol priority
  16. · Alex's posts · Alex Getman
    Excerpt
    🚨 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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    ⚡️ Muse Spark takes 3rd place on the AA Intelligence Index, tying Claude Fable 5 and beating GPT-5.6 Sol
  19. · Alex's posts · Alex Getman
    Excerpt
    🚨 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. · Alex's posts · Alex Getman
    Excerpt
    ⚡️ 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. · Alex's posts · Alex Getman
    Excerpt
    ⚡️ 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. · Alex's posts · Alex Getman
    Excerpt
    ⚡️ 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. · Alex's posts · Alex Getman
    Excerpt
    🎉 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. · Alex's posts · Alex Getman
    Excerpt
    ⚡️ 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. · Alex's posts · Alex Getman
    Excerpt
    ⚡️ 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. · Alex's posts · Alex Getman
    Excerpt
    ⚡️ 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. · Alex's posts · Alex Getman
    Excerpt
    🚨 OPENAI RESET CODEX LIMITS • GPT-5.6 Sol now drains your limits 18% slower. • The five-hour limit returns tomorrow.
  28. · Alex's posts · Alex Getman
    Excerpt
    🚨 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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    CODEX GIVEAWAY $100 in API credits Post about GPT-5.6 Sol on Twitter. First 10K get the bag.
  34. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    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. · Alex's posts · Alex Getman
    Excerpt
    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

Alex's take

Signals 2

OpenAISighting

GPT 5.6 Sol

Confirmed22 evidence links

Alex's posts 10

Specifications

GPT-5.6 Sol is a model from OpenAI, released on 9 Jul 2026. The context window is 1,050,000 tokens. The maximum output is 128,000 tokens. Pricing is $4.00 per million input tokens and $20.00 per million output tokens.

Context window
1,050,000
Max output
128,000
Input
text · image
Output
text
Input · USD / 1M tokens
$4.00
Output · USD / 1M tokens
$20.00
Cached input · USD / 1M tokens
$0.40

Questions

What is the context window of GPT-5.6 Sol?

The context window is 1,050,000 tokens. The maximum output is 128,000 tokens.

How much does GPT-5.6 Sol cost?

Pricing is $4.00 per million input tokens and $20.00 per million output tokens. Cached input is $0.40 per million.

When was GPT-5.6 Sol released?

On 9 Jul 2026.