{ "schema_version": "ctz-coding-harness-v1", "generated_at": "2026-06-30T13:54:46Z", "source_report_generated_at": "2026-06-30T13:52:11Z", "source_health": { "target_count": 40, "success_count": 40, "error_count": 0, "error_ids": [], "used_source_count": 32, "dynamic_shell_sources": [ "zai_subscribe" ] }, "plans": [ { "id": "opencode_go", "product": "OpenCode Go", "vendor": "OpenCode", "harness_family": "model_pool_harness_plan", "price_label": "$10/mo", "entry_price_usd": 10, "serious_use_price": null, "observed_price_points_usd": [ 5, 10, 12, 30, 60 ], "capacity_unit": "dollar-value usage pool with model-specific request estimates", "capacity_amount": "$12 per 5 hours, $30 weekly, and $60 monthly usage pool; request count depends on chosen model.", "reset_window": "5-hour, weekly, monthly", "over_limit_behavior": "continue with free models or Zen balance if enabled", "model_access": "Low-cost coding model pool with explicit per-model request estimates.", "highlighted_models": [ "GLM-5.2", "GLM-5.1", "Kimi K2.7 Code", "Qwen3.7 Max", "MiniMax M3", "DeepSeek V4 Pro" ], "model_multiplier": "higher-cost models like GLM-5.2 allow fewer requests; cheaper models allow more requests", "shared_limit_scope": "Go usage pool for supported model access through OpenCode-compatible flows", "harness_surface": [ "OpenCode", "TUI", "compatible agents through API key", "Zen balance fallback" ], "capacity_translation": "GLM-5.2: 880 requests/5h, 2,150/week, 4,300/month; cheaper models can be much higher", "risk_tags": [ "model_pool_not_frontier_default", "capacity_depends_on_model", "usage_limits_may_change", "zen_balance_fallback" ], "capacity_clarity_score": 92, "cost_safety_score": 82, "heavy_coding_fit_score": 68, "best_for": "Low-cost coding volume on a transparent model pool with request estimates.", "watch_out": "It is a model-pool plan, not a default GPT/Claude frontier subscription.", "upgrade_when": "GLM-5.2 or similar higher-cost models hit the 5-hour or monthly pool.", "source_confidence": "high", "source_urls": [ "https://opencode.ai/go", "https://opencode.ai/docs/go/", "https://opencode.ai/docs/zen/" ] }, { "id": "zai_glm_coding_plan", "product": "GLM Coding Plan", "vendor": "Z.AI", "harness_family": "model_pool_harness_plan", "price_label": "$18/mo", "entry_price_usd": 18, "serious_use_price": null, "observed_price_points_usd": [ 18 ], "capacity_unit": "prompt quota", "capacity_amount": "Lite, Pro, and Max plans expose approximate 5-hour and weekly prompt limits.", "reset_window": "5-hour dynamic refresh and weekly reset; separate monthly MCP quota", "over_limit_behavior": "wait for the next 5-hour or weekly reset; docs say the system will not deduct from account balance for supported tool usage when plan quota runs out", "model_access": "GLM coding-focused model access for supported tools.", "highlighted_models": [ "GLM-5.2", "GLM-5-Turbo", "GLM-4.7" ], "model_multiplier": "GLM-5.2 and GLM-5-Turbo can deduct at higher peak-hour multipliers; docs recommend GLM-5.2 for complex tasks and GLM-4.7 for routine work to preserve quota", "shared_limit_scope": "strictly limited to officially supported tools and products", "harness_surface": [ "Claude Code", "Cline", "OpenCode", "officially supported tools and products" ], "capacity_translation": "Lite: about 80 prompts/5h and 400/week; Pro: 400/5h and 2,000/week; Max: 1,600/5h and 8,000/week; FAQ also notes each prompt may invoke 15-20 model calls", "risk_tags": [ "tool_locked", "subscribe_page_dynamic_shell", "prompt_estimates_not_guarantees", "separate_mcp_limits" ], "capacity_clarity_score": 80, "cost_safety_score": 76, "heavy_coding_fit_score": 64, "best_for": "GLM-heavy workflows inside supported tools such as Claude Code, Cline, and OpenCode.", "watch_out": "Tool lock, prompt estimates, peak multipliers, and separate MCP limits matter.", "upgrade_when": "The Lite prompt pool is too small or Pro/Max weekly limits are needed.", "source_confidence": "medium_high", "source_urls": [ "https://docs.z.ai/devpack/overview", "https://docs.z.ai/devpack/latest-model", "https://docs.z.ai/devpack/faq", "https://z.ai/subscribe" ] }, { "id": "gemini_code_assist", "product": "Gemini Code Assist", "vendor": "Google", "harness_family": "ide_and_cloud_harness", "price_label": "$22.8/mo", "entry_price_usd": 22.8, "serious_use_price": null, "observed_price_points_usd": [ 19, 22.8, 45, 54 ], "capacity_unit": "per-user license plus request quotas", "capacity_amount": "Standard and Enterprise are licensed per user per month; Gemini Code Assist quota docs publish agent mode/Gemini CLI daily request limits by license.", "reset_window": "daily request quota for Gemini CLI/agent features; license billed monthly or annually", "over_limit_behavior": "requests stop for agent mode/Gemini CLI once the daily maximum is reached until reset; higher limits require program, Pro/Ultra, Standard, Enterprise, or quota adjustment path", "model_access": "Gemini 3 and Gemini Code Assist model access for Standard and Enterprise subscribers.", "highlighted_models": [ "Gemini 3", "Gemini CLI" ], "model_multiplier": "not exposed as per-model credit burn in pricing page; Gemini 3 access called out for Standard and Enterprise", "shared_limit_scope": "Gemini CLI and Code Assist agent usage are related surfaces under the license", "harness_surface": [ "supported IDEs", "JetBrains", "Gemini CLI", "Google Cloud surfaces", "agent mode preview", "MCP ecosystem tools" ], "capacity_translation": "Standard: 1,500 requests/user/day; Enterprise: 2,000 requests/user/day; GitHub app PR-review quota is separate", "risk_tags": [ "license_based", "daily_request_cap", "github_pr_review_separate_quota", "region_or_cloud_context_possible" ], "capacity_clarity_score": 86, "cost_safety_score": 78, "heavy_coding_fit_score": 70, "best_for": "Google Cloud and enterprise teams that value daily request quotas and governed IDE/CLI access.", "watch_out": "Agent mode and Gemini CLI daily requests stop after quota reset thresholds are reached.", "upgrade_when": "Standard daily requests are too tight or Google Cloud governance becomes a requirement.", "source_confidence": "high", "source_urls": [ "https://codeassist.google/products/business", "https://developers.google.com/gemini-code-assist/resources/quotas", "https://developers.google.com/program/plans-and-pricing" ] }, { "id": "github_copilot", "product": "GitHub Copilot", "vendor": "GitHub", "harness_family": "ide_and_github_harness", "price_label": "$10/mo -> $39/mo -> $100/mo", "entry_price_usd": 10, "serious_use_price": 100, "observed_price_points_usd": [ 0, 10, 15, 39, 70, 100, 200 ], "capacity_unit": "monthly GitHub AI credits plus unlimited paid completions", "capacity_amount": "Pro includes $15 monthly credits, Pro+ $70, Max $200; free tier has limited completions and chat requests.", "reset_window": "monthly credit cycle", "over_limit_behavior": "Individuals can move up tiers or use paid usage where available; Business and Enterprise depend on admin budget policy and may pause until next cycle.", "model_access": "OpenAI, Anthropic, Gemini, and third-party agent access varies by plan.", "highlighted_models": [ "GPT-5.5", "Claude Opus 4.8", "Claude Sonnet 4.6", "Gemini 3" ], "model_multiplier": "frontier-model agent sessions and longer sessions consume more credits", "shared_limit_scope": "chat, agents, CLI, Spaces, and Spark consume credits; code completions and next edit suggestions do not consume credits on paid plans", "harness_surface": [ "VS Code", "Visual Studio", "JetBrains", "GitHub.com", "GitHub CLI", "mobile", "agents", "code review" ], "capacity_translation": "credit-denominated; exact sessions/month depends on model and feature mix", "risk_tags": [ "credit_metered_agents", "admin_budget_controls", "feature_specific_metering", "model_selection_multiplier" ], "capacity_clarity_score": 74, "cost_safety_score": 70, "heavy_coding_fit_score": 72, "best_for": "GitHub-centered teams that want IDE help, agents, CLI, review, and admin controls.", "watch_out": "Agent, chat, CLI, Spaces, and Spark consume AI credits while completions do not.", "upgrade_when": "Premium model work or agent sessions consume the included credit pool quickly.", "source_confidence": "high", "source_urls": [ "https://github.com/features/copilot/plans", "https://docs.github.com/en/billing/concepts/product-billing/github-copilot-licenses" ] }, { "id": "cursor", "product": "Cursor", "vendor": "Anysphere", "harness_family": "ide_agent_harness", "price_label": "$20/mo -> $60/mo -> $200/mo", "entry_price_usd": 20, "serious_use_price": 200, "observed_price_points_usd": [ 0, 20, 40, 60, 200 ], "capacity_unit": "included model usage plus on-demand usage", "capacity_amount": "Plans include a set amount of model usage; Cursor recommends Pro+ for daily agent users and Ultra for agent power users. Official tier update says Pro moved from request limits to compute limits and Ultra has 20x more usage than Pro.", "reset_window": "monthly billing cycle; exact included usage units not exposed by static scrape", "over_limit_behavior": "on-demand usage after included amount is consumed, billed in arrears", "model_access": "Access to frontier models with plan-dependent included usage.", "highlighted_models": [ "frontier model pool" ], "model_multiplier": "frontier model and agent usage intensity changes burn rate; Pro compute is described in API-price model inference terms", "shared_limit_scope": "usage is tied to individual or team/pooled plan depending on subscription", "harness_surface": [ "Cursor IDE", "agent", "tab completions", "cloud agents", "automations", "rules", "skills", "plugins" ], "capacity_translation": "Pro includes at least $20 of model inference at API prices per month; Ultra is positioned as 20x more usage than Pro", "risk_tags": [ "usage_based_overage", "exact_quota_opaque", "team_pooled_usage_possible", "dynamic_docs_gap" ], "capacity_clarity_score": 66, "cost_safety_score": 60, "heavy_coding_fit_score": 84, "best_for": "IDE-first developers who want agent work fused into editing, tabs, rules, and cloud agents.", "watch_out": "Included model usage can roll into on-demand billing after the plan allowance is consumed.", "upgrade_when": "Daily agent usage needs Pro+ or Ultra predictability.", "source_confidence": "high_price_medium_capacity", "source_urls": [ "https://cursor.com/pricing", "https://cursor.com/docs/models-and-pricing", "https://cursor.com/blog/new-tier" ] }, { "id": "openai_codex", "product": "OpenAI Codex", "vendor": "OpenAI", "harness_family": "official_agent_harness", "price_label": "$8/mo -> $100/mo", "entry_price_usd": 8, "serious_use_price": 100, "observed_price_points_usd": [ 0, 8, 20, 25, 100 ], "capacity_unit": "token-based Codex credits plus plan rate limits", "capacity_amount": "Codex is included in ChatGPT plans; Plus is positioned for a few focused coding sessions each week, Pro provides 5x or 20x higher limits than Plus, and the rate card maps usage to credits per 1M input/cache/output tokens.", "reset_window": "plan limits vary", "over_limit_behavior": "Use ChatGPT credits or run local tasks with an API key billed at API pricing.", "model_access": "Latest Codex models exposed through ChatGPT/Codex plan access.", "highlighted_models": [ "GPT-5.5", "GPT-5.4", "GPT-5.4 mini", "GPT-5.3" ], "model_multiplier": "credit consumption varies by model, input/cache/output token mix, number of instances, automations, and fast mode", "shared_limit_scope": "bundled with broader ChatGPT plan access", "harness_surface": [ "web", "CLI", "IDE extension", "iOS", "cloud integrations", "local tasks via API key" ], "capacity_translation": "not directly published as requests/month; needs workload translation from credit rate card and observed sessions", "risk_tags": [ "shared_chatgpt_plan", "credit_rate_multiplier", "api_key_overage_possible", "opaque_session_capacity" ], "capacity_clarity_score": 52, "cost_safety_score": 58, "heavy_coding_fit_score": 76, "best_for": "ChatGPT-native developers who want web, CLI, IDE, and cloud tasks in one subscription.", "watch_out": "Token-credit burn varies by model, output, fast mode, and API-key overage.", "upgrade_when": "Plus sessions are not enough or Codex becomes a daily cloud-task workflow.", "source_confidence": "high_source_medium_capacity", "source_urls": [ "https://developers.openai.com/codex/pricing", "https://openai.com/chatgpt/pricing/", "https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan", "https://help.openai.com/en/articles/20001106-codex-rate-card" ] }, { "id": "claude_code", "product": "Claude Code", "vendor": "Anthropic", "harness_family": "official_agent_harness", "price_label": "$20/mo -> $100/mo or $200/mo", "entry_price_usd": 20, "serious_use_price": [ 100, 200 ], "observed_price_points_usd": [ 0, 17, 20, 100, 125, 200 ], "capacity_unit": "shared Claude usage limits", "capacity_amount": "Pro and Max subscriptions cover Claude Code and Claude app usage under shared limits; Max gives larger capacity but not unlimited usage.", "reset_window": "usage windows vary by plan and remaining capacity", "over_limit_behavior": "Wait for reset, upgrade to Max 5x or Max 20x, or explicitly consent to API-credit usage billed at API rates.", "model_access": "Claude models through Pro/Max and API-credit fallback.", "highlighted_models": [ "Claude Opus", "Claude Sonnet", "Claude Haiku" ], "model_multiplier": "Opus costs several times more per turn than Sonnet, and Sonnet more than Haiku; long sessions and large context also drain limits faster", "shared_limit_scope": "Claude Code and Claude app activity share usage limits", "harness_surface": [ "terminal", "VS Code", "Cursor and VS Code forks", "JetBrains", "Claude web", "desktop", "mobile" ], "capacity_translation": "not published as fixed requests/month; needs measured workload samples", "risk_tags": [ "shared_claude_limits", "api_credit_opt_in", "opaque_capacity", "model_complexity_multiplier" ], "capacity_clarity_score": 50, "cost_safety_score": 56, "heavy_coding_fit_score": 82, "best_for": "Deep repository reasoning, complex refactors, and planning-heavy coding work.", "watch_out": "Claude app and Claude Code share usage limits; Opus drains quota much faster.", "upgrade_when": "Sonnet work regularly hits limits or Opus planning becomes part of the routine.", "source_confidence": "high_source_medium_capacity", "source_urls": [ "https://www.anthropic.com/pricing", "https://support.anthropic.com/en/articles/11049762-choosing-a-claude-ai-plan", "https://support.anthropic.com/en/articles/11145838-using-claude-code-with-your-pro-or-max-plan", "https://docs.anthropic.com/en/docs/claude-code/costs", "https://support.anthropic.com/en/articles/14552983-models-usage-and-limits-in-claude-code" ] }, { "id": "devin_windsurf", "product": "Devin Desktop / Windsurf", "vendor": "Cognition", "harness_family": "agentic_workspace_harness", "price_label": "$20/mo -> $200/mo", "entry_price_usd": 20, "serious_use_price": 200, "observed_price_points_usd": [ 0, 20, 40, 80, 200 ], "capacity_unit": "daily and weekly usage allowance", "capacity_amount": "Paid plans include usage allowances that refresh daily and weekly; Max has significantly higher quotas.", "reset_window": "daily and weekly", "over_limit_behavior": "purchase extra usage consumed at API pricing", "model_access": "SWE models, leading open source models, and frontier models from OpenAI, Claude, and Gemini on higher quotas.", "highlighted_models": [ "SWE 1.6", "OpenAI frontier", "Claude frontier", "Gemini frontier" ], "model_multiplier": "cost per message varies by model, task size, complexity, reasoning, and token usage", "shared_limit_scope": "plan allowance covers agent usage; Teams includes per-seat quota and Devin Desktop access", "harness_surface": [ "Devin Desktop", "Windsurf", "agents", "inline edits", "tab completions", "teamspace", "Slack and Teams integrations" ], "capacity_translation": "not published as fixed requests/month; measured message/task samples needed", "risk_tags": [ "usage_allowance_opaque", "api_pricing_overage", "model_complexity_multiplier", "daily_weekly_quota" ], "capacity_clarity_score": 54, "cost_safety_score": 55, "heavy_coding_fit_score": 78, "best_for": "Agentic workspace users who want inline edits, tab completion, teamspace, and higher automation quotas.", "watch_out": "Task complexity, model choice, reasoning, and token volume make capacity hard to predict.", "upgrade_when": "Daily/weekly quota runs out or extra usage at API pricing becomes frequent.", "source_confidence": "high_price_medium_capacity", "source_urls": [ "https://devin.ai/pricing/", "https://windsurf.com/pricing", "https://docs.windsurf.com/windsurf/accounts/quota" ] } ], "excluded_app_builders": [ { "id": "v0", "product": "v0", "vendor": "Vercel", "reason": "AI app builder: prompt-to-app credits, hosting/deploy workflow, and team app generation economics are not coding harness capacity.", "source_urls": [ "https://v0.app/pricing" ] }, { "id": "bolt", "product": "Bolt", "vendor": "StackBlitz", "reason": "AI app builder: token-based app generation workflow should remain a separate lane.", "source_urls": [ "https://bolt.new/pricing" ] }, { "id": "replit_agent", "product": "Replit Agent", "vendor": "Replit", "reason": "AI app builder/workspace: effort-based app agent billing does not compare cleanly with coding harness subscriptions.", "source_urls": [ "https://replit.com/pricing", "https://docs.replit.com/billing/ai-billing" ] }, { "id": "lovable", "product": "Lovable", "vendor": "Lovable", "reason": "AI app builder: build credits and app workflow belong outside the main coding harness ranking.", "source_urls": [ "https://lovable.dev/pricing" ] } ], "scoring_notes": [ "Quota Clarity estimates how well the published plan can be translated into real requests, prompts, credits, reset windows, and model multipliers.", "Cost Safety estimates how explicit the over-limit path is, including wait/reset, downgrade, paid overage, API billing, admin budgets, and surprise-spend risk.", "Heavy Coding Fit estimates suitability for daily agentic coding, considering agent surfaces, model access, quota depth, workflow integration, and team/cloud automation." ] }