Multi-Model Workflow · As of 2026-11-18

Multi-Model Coding API Without Subscription Lock-In: One Key for Claude Code, Codex & Cline

As of 2026-11-18, developers are increasingly hitting weekly subscription caps mid-sprint, preparing for OpenAI's December 2026 shared ChatGPT and Codex usage pool, or facing cross-border card and account verification friction. Instead of juggling multiple $20 to $200 monthly subscriptions, connect your local CLI and IDE tools to a per-token API workflow: one QCode cr_ API key powers Claude Code (https://api.qcode.cc/api), Codex CLI (https://api.qcode.cc/openai), and Cline, OpenCode, or WorkBuddy (https://api.qcode.cc/openai/v1) across claude-fable-5-2, claude-opus-5-5, gpt-6.2-sol, gpt-6.2-astra, qwen4-max, deepseek-v4.1-flash, and kimi-k3.1-pro.

Updated 2026-11-18

One key for Claude, GPT and Chinese models — put what this page describes into practice.

#One cr_ API key#3 native endpoints#Per-token billing#Flexible payments & packs

Why engineering teams replace stacked subscriptions with one API key

2026-12

Shared ChatGPT + Codex pool and weekly caps

OpenAI Help Center (20001625) announced that starting in December 2026, Plus, Pro, and Business usage allowances are shared between ChatGPT and Codex; Anthropic's S-1 (2026-11-10) also notes usage and rate limits may change at any time, making fixed plans easy to exhaust during heavy agent runs.

3 Endpoints

Claude Code, Codex CLI, and Cline on one key

Route the same QCode cr_ key by tool protocol: set https://api.qcode.cc/api in Claude Code, https://api.qcode.cc/openai in Codex CLI (wire_api = "responses"), and https://api.qcode.cc/openai/v1 in Cline, OpenCode, WorkBuddy, and the OpenAI SDK.

7 Flagships

Call US and Chinese coding flagships side by side

Without maintaining separate vendor balances, one key calls claude-fable-5-2, claude-opus-5-5, gpt-6.2-sol, gpt-6.2-astra, qwen4-max, deepseek-v4.1-flash, and kimi-k3.1-pro (plus glm-5.3), including instant in-session switches via /model in Claude Code.

7-Day Packs

Flexible payment rails and short-term quota control

Top up via Alipay, WeChat Pay, international Visa / Mastercard credit and debit cards, Apple Pay, Google Pay, or USDT / USDC / BTC / ETH crypto; choose a 30-day Personal Plan or grab a 7-day Pay-as-you-go Pack or booster pack when sprints spike.

Where fixed monthly subscription lock-in breaks down in November 2026

As of 2026-11-18, modern AI engineering workflows rarely rely on a single model family: teams pair claude-fable-5-2 or claude-opus-5-5 for multi-file architecture with the 2026-11-17 released gpt-6.2-sol and gpt-6.2-astra for Codex terminal execution and deep reasoning, plus the 2026-11-16 qwen4-max, 2026-11-13 updated deepseek-v4.1-flash, or kimi-k3.1-pro for 1M-token context analysis and high-concurrency sub-agents. Stacking individual $20, $100, and $200 monthly subscriptions creates three structural bottlenecks: ① Shared and rolling quota ceilings — OpenAI announced that starting in December 2026, Plus, Pro, and Business plans share one allowance pool across ChatGPT and Codex, while Anthropic's Pro, Max, and Ultra plans enforce weekly ceilings and rolling windows (only verified Ultra accounts remove the 5-hour session cap), leaving developers waiting for resets or buying extra credits mid-release; ② Disclosed rate-limit volatility — Anthropic's 2026-11-10 Form S-1/A reveals that 80% ($13.064B) of H1 2026 revenue comes from metered API consumption and explicitly warns in Risk Factors that usage limits and rate limits may be modified at any time; ③ Cross-border billing and verification friction — Anthropic's 2026-11-12 Supported Regions Policy (enforcing 3 dimensions: physical location, entity registration/HQ, and >50% ownership/control), Persona photo-ID verification (Help Center 14328960), and no-refund rule on policy terminations (Help Center 12386328) make personal plans fragile for distributed teams. Pointing local CLI tools at a per-token multi-model API key removes single-subscription lock-in cleanly. See /pricing and /pay-as-you-go for plan details, and /usdt-buy-claude-code-credits for crypto top-ups.

Mid-November 2026 flagship lineup: official vendor API rates and specs

Between 2026-11-05 and 2026-11-17, every major coding model family shipped a new flagship. Here are the official vendor API rates per 1M tokens and core specs: ① Anthropic claude-fable-5-2 (released 2026-11-09, default Fable model since Claude Code v2.1.321): 1M context, 128K max output, official Anthropic rate $8 input / $40 output; while active Max ($100/month), Ultra ($200/month), and Team plans added non-rollover monthly platform API credits on 2026-11-17 (claimable after 7 days, Help Center 17154008), they remain tied to a single vendor. ② OpenAI gpt-6.2-sol and gpt-6.2-astra (released 2026-11-17): 1,050,000 context, 922,000 max input, 128,000 max output (rejecting reasoning_effort="none"); at Standard speed up to 272K input tokens, the official OpenAI rate for gpt-6.2-sol is $2 input, $0.10 cached input, and $10 output (within 0.5 points of Astra on FrontierCode 2.0), while gpt-6.2-astra is $8 input, $0.40 cached input, and $40 output (20% lower than gpt-6-astra's $10 / $50; requests over 272K input tokens are billed at 2x input and 1.5x output). ③ Alibaba Cloud Model Studio qwen4-max (general availability 2026-11-16): 2.4T total / 71B active MoE, flat pricing across all 1,000,000 tokens with no long-context doubling at the official Alibaba Cloud rate of ¥10 / ¥30 in mainland China or $1.50 / $4.50 on the Singapore international node, scoring 52.69 on Work360 at $1.67 average task cost. ④ DeepSeek deepseek-v4.1-flash (updated 2026-11-13 to DeepSeek-V4.1-Flash-1017): 1M context, 384K max output, 77.0 on DeepSWE v1.1, official DeepSeek rate $0.15 / $0.60 off-peak and $0.30 / $1.20 peak. ⑤ Moonshot kimi-k3.1-pro (launched 2026-11-05): 1M context, 131,072 max output, native vision, official Moonshot rate $2.00 input, $10.00 output, $0.20 cache hit.

Key November–December 2026 model and subscription milestones

2026-11-12

Anthropic's 3-dimension Supported Regions Policy and updated Usage Policy take effect; QCode lists claude-fable-5-2 and kimi-k3.1-pro on /models.

2026-11-16

Alibaba Cloud Model Studio launches general-availability pay-as-you-go qwen4-max (flat 1M rate at ¥10 / ¥30 or $1.50 / $4.50), following DeepSeek's 2026-11-13 DeepSeek-V4.1-Flash-1017 update.

2026-11-17

OpenAI launches gpt-6.2-astra ($8 / $40) and gpt-6.2-sol ($2 / $10) alongside its notice that ChatGPT and Codex share subscription allowances starting in December 2026; all new flagships are callable on QCode as of 2026-11-18.

Confirmed official policies vs compliance and billing boundaries

Confirmed (official vendor docs and live QCode specifications)

Verified as of 2026-11-18: ① OpenAI Help Center (20001625): starting in December 2026, Plus, Pro, and Business usage allowances are shared between ChatGPT and Codex (Instant chat does not consume the shared pool); gpt-6.2-astra and gpt-6.2-sol launched on 2026-11-17 and reject reasoning_effort="none". ② Anthropic official filings and docs: Form S-1/A (2026-11-10) discloses an 80% H1 2026 API revenue share ($13.064B) and notes usage or rate limits may change; the 3-dimension Supported Regions Policy and updated Usage Policy took effect on 2026-11-12; Help Center articles (14328960, 12386328, 17154008) confirm Persona ID verification, no refund for policy-terminated subscriptions, and monthly platform API credits ($100 on active Max, $200 on active Ultra) starting 2026-11-17 (claimable after 7 days). ③ Chinese flagship docs: qwen4-max launched 2026-11-16 (2.4T/71B MoE, flat 1M rate ¥10 / ¥30 or $1.50 / $4.50); deepseek-v4.1-flash updated 2026-11-13 to build 1017 (DeepSWE v1.1 at 77.0, Terminal-Bench 4.0 at 42.3); kimi-k3.1-pro launched 2026-11-05 (1M / 131,072). ④ QCode endpoints and payment rails: one cr_ key works across https://api.qcode.cc/api, https://api.qcode.cc/openai, and https://api.qcode.cc/openai/v1; payments support Alipay, WeChat Pay, Visa / Mastercard, Apple Pay, Google Pay, and USDT / USDC / BTC / ETH; billed per token with per-model rates on /models.

Compliance boundaries and unannounced details

As of 2026-11-18: ① OpenAI has not announced the exact calendar day in December 2026 when the shared ChatGPT and Codex allowance pool takes effect (stating only December 2026). ② QCode serves only the production model IDs listed on /models and does not offer gpt-6.2-mini, gpt-6.2-terra, gpt-6.2-luna, gpt-6.2-codex, qwen4-plus, qwen4-flash, qwen4-coder, or qwen4-max-preview. ③ Compliance note: this guide covers only standard, legitimate API key configurations for local CLI and IDE tools across documented endpoints, and provides zero advice on bypassing regional restrictions, switching network nodes, or spoofing identity verification.

Stacking multiple monthly subscriptions vs one QCode multi-model API key

Stacking monthly plans: $120 to $400+ fixed overhead, weekly caps, and siloed balances

Maintaining separate Claude Pro/Max/Ultra, ChatGPT Plus/Pro, and Chinese coding plans easily costs $120 to $400+ every month; unused quota expires during quiet weeks, while heavy release weeks run into 5-hour session windows, weekly caps, or the December 2026 shared ChatGPT+Codex pool with balances locked inside separate vendor silos.

One QCode cr_ API key: per-token billing, shared balance across tools, and instant model switching

With a unified QCode API key, Claude Code, Codex CLI, Cline, OpenCode, and WorkBuddy draw from one token balance: route complex architecture to claude-fable-5-2, claude-opus-5-5, or gpt-6.2-astra, everyday coding to gpt-6.2-sol or qwen4-max, parallel sub-agents to deepseek-v4.1-flash, and visual/long-context tasks to kimi-k3.1-pro, paying strictly per token used.

Copy-paste setup for Claude Code, Codex CLI, and Cline

① Claude Code (Anthropic Messages protocol): set ANTHROPIC_BASE_URL=https://api.qcode.cc/api and ANTHROPIC_AUTH_TOKEN to your cr_ key in your shell or ~/.claude/settings.json env block; run /model claude-fable-5-2 or /model claude-opus-5-5 for Claude flagships, or type /model qwen4-max, /model deepseek-v4.1-flash, /model kimi-k3.1-pro, or /model glm-5.3 in the same session to switch to Chinese flagships (note that Claude models only use https://api.qcode.cc/api). ② Codex CLI (OpenAI Responses protocol): in ~/.codex/config.toml set wire_api = "responses" and base_url = "https://api.qcode.cc/openai" (no /v1 suffix; use "https://asia.qcode.cc/openai" in mainland China or "https://us.qcode.cc/openai" in North America and Europe), set model to "gpt-6.2-sol" or "gpt-6.2-astra", and set model_reasoning_effort to "high" or "xhigh" (the 6.2 family rejects "none"). ③ Cline / OpenCode / WorkBuddy / OpenAI SDK: choose OpenAI Compatible mode, set Base URL to https://api.qcode.cc/openai/v1 (or https://asia.qcode.cc/openai/v1 in mainland China, with no trailing slash), enter the same cr_ key, and set the model ID to gpt-6.2-sol, gpt-6.2-astra, qwen4-max, deepseek-v4.1-flash, kimi-k3.1-pro, or glm-5.3.

Choose Pay-as-you-go Packs or Personal Plans on QCode

QCode supports claude-fable-5-2, claude-opus-5-5, claude-sonnet-5-5, gpt-6.2-sol, gpt-6.2-astra, qwen4-max, deepseek-v4.1-flash, kimi-k3.1-pro, and glm-5.3 with a single cr_ key across all three developer endpoints. Billing is per token, with per-model rates on /models. Not ready for a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days; if you already have a Personal Plan, add a booster pack when you need more quota.

Frequently asked questions

Can one QCode cr_ key really run Claude Code, Codex CLI, and Cline at the same time?

Yes. The same cr_ key works across all three protocol endpoints: Claude Code uses ANTHROPIC_BASE_URL=https://api.qcode.cc/api (supporting all Claude models plus in-session switches like /model qwen4-max and /model glm-5.3); Codex CLI uses base_url = "https://api.qcode.cc/openai" in config.toml (no /v1 suffix, calling gpt-6.2-sol, gpt-6.2-astra, and more); Cline, OpenCode, and WorkBuddy use https://api.qcode.cc/openai/v1.

What changes in December 2026 for ChatGPT and Codex subscription usage limits?

Per OpenAI Help Center article 20001625, starting in December 2026, ChatGPT Plus, Pro, and Business plans share a single usage allowance pool between ChatGPT and Codex (Instant chat does not draw from that pool; OpenAI has not yet announced the exact December date). Once the shared cap is reached, users must wait for the cycle reset or use per-token API billing.

How do I buy short-term coding quota on QCode without a recurring monthly subscription?

You can buy a Pay-as-you-go Pack directly in the dashboard: it gives you a one-time quota valid for 7 days, deducted per token with no monthly commitment; if you already hold a 30-day Personal Plan and hit a busy sprint, you can add a booster pack anytime. Live per-model token rates are listed on /models.

Can I top up with Alipay, WeChat Pay, Apple Pay, or USDT if I do not have a USD card?

Yes. The QCode dashboard separates top-ups into CNY and USD tabs: the CNY tab accepts Alipay and WeChat Pay (marked 0% platform fee); the USD tab accepts international Visa / Mastercard credit and debit cards ($5 to $2,000 per transaction, credited immediately), Apple Pay, Google Pay, and cryptocurrency including USDT, USDC, BTC, and ETH (supporting Tron TRC20, Solana, Arbitrum One, Polygon, Ethereum, TON, and more at 0% platform fee, credited after manual on-chain review in about 1 day).

What parameter rule should I watch when calling gpt-6.2-sol or gpt-6.2-astra in Codex CLI?

OpenAI's official docs state that the 2026-11-17 released gpt-6.2-astra and gpt-6.2-sol (both featuring a 1,050,000-token context and 128,000 max output) explicitly reject requests with reasoning_effort="none". Keep reasoning effort set to high or xhigh in Codex CLI or API payloads, and set Codex CLI's base_url to https://api.qcode.cc/openai without a trailing /v1.

How should I mix Claude 5.2, GPT-6.2, and Qwen4-Max to optimize my coding budget?

A practical engineering split is: ① run core architecture and hard debugging on claude-fable-5-2 (official Anthropic rate $8 / $40) or gpt-6.2-astra (official OpenAI rate $8 / $40); ② run daily coding and PR reviews on gpt-6.2-sol (official OpenAI rate $2 / $10, within 0.5 points of Astra on FrontierCode 2.0) or qwen4-max (official Alibaba Cloud rate ¥10 / ¥30 or $1.50 / $4.50 flat across 1M tokens, averaging $1.67 per task on Work360); ③ run high-concurrency test loops on deepseek-v4.1-flash (official DeepSeek off-peak rate $0.15 / $0.60).

Sources

OpenAI official: gpt-6.2-astra and gpt-6.2-sol model pages (2026-11-17) and Help Center shared usage notice (20001625); Anthropic official: Form S-1/A (2026-11-10), Supported Regions Policy and Usage Policy (2026-11-12), and Help Center (14328960, 12386328, 17154008); Alibaba Cloud Model Studio, DeepSeek, and Moonshot official pricing docs; QCode: qcode.cc/models, qcode.cc/pricing, qcode.cc/pay-as-you-go, and docs.qcode.cc. All checked on 2026-11-18.

Skip subscription lock-in and run seven coding flagships with one key

Create a QCode cr_ API key to run claude-fable-5-2, gpt-6.2-sol, and qwen4-max in Claude Code, Codex CLI, and Cline — billed per token with rates on /models.

Not sure where to start?

Try it for a week

Small Pack: $4 for $30 of quota, valid 7 days

Use it every day

Starter: $9.00/month with $5 of quota every day — just $0.060 per $1 of quota, vs about $0.113 for the Medium Pack