{"slug": "vibe-code-tours-student-setup-linux-macos-wsl-nvm-node22-uv-python3-12-git-code", "title": "Vibe Code Tours — student setup (Linux/macOS/WSL): nvm+Node22, uv+Python3.12, git, Claude Code + verify", "summary": "A developer has published a verified pricing and setup guide for AI coding tools, comparing free and paid options for students using Linux, macOS, or WSL. The guide recommends Google Gemini CLI for zero-cost, no-card-required access with 60 requests per minute, or opencode with a Gemini API key for multi-provider flexibility, while noting Claude Code as the top-quality paid option at $5–10 per student per month. The post includes detailed per-model pricing tables for providers including Anthropic, OpenAI, Google, DeepSeek, and others, with all figures verified as of early-mid 2026.", "body_md": "Quick reference for picking a provider/CLI on day 1. **All numbers verified early-mid 2026 — confirm live before you depend on them.** Pricing and quotas change frequently (especially Chinese providers and free-tier limits).\n\n| Goal | Pick | Why |\n|---|---|---|\nZero cost, zero card |\nGemini CLI with personal Google account |\nGenerous free tier, no card required, 60 req/min |\n| Zero cost + multi-provider | opencode + Gemini API key (AI Studio free tier) |\nModel-agnostic CLI; swap providers later without changing tool |\n| Best quality, paid | Claude Code with Anthropic API or Bedrock | Top coding quality; budget $5–10/student/month is realistic |\n| Cheapest paid bulk | DeepSeek V4 Flash via opencode | $0.14/$0.28 per M tokens — 20× cheaper than Claude |\n\nIf your region blocks `claude.ai`\n\n, **skip Claude Code on day 1** — use Gemini CLI or opencode instead. We'll add the regional fallback later.\n\n**Provider:** Google Gemini Code Assist (consumer tier).**Auth:** browser sign-in with a personal Google account on first run.**Quota:** up to**60 requests/minute**,**~1,000 requests/day** on Flash; cap drops to ~100/day on Pro-class models.**Cost:****$0**. No card. May log prompts for training (read the consent screen).** Install:**`npm i -g @google/gemini-cli`\n\n→`gemini`\n\n.\n\n**Provider:**`ai.google.dev`\n\n— separate from Gemini CLI; gives you an API key you can wire into opencode/LiteLLM/anything.**Free quotas**(per-day, per-model, no card):** Gemini 2.5 Pro**— ~100 requests/day · 5 req/min** Gemini 2.5 Flash**— ~250 requests/day · 10 req/min** Gemini 2.5 Flash-Lite**— ~1,500 requests/day · 15 req/min\n\n**Cost:****$0** within the quota; over-quota = 429 until reset.**Get a key:**[https://aistudio.google.com/app/apikey](https://aistudio.google.com/app/apikey)\n\n**Free BYOK:****1,000,000 requests/month** routed through OpenRouter to your own provider keys, then 5% surcharge.real free hosted variants (Qwen3-Coder, DeepSeek, GLM, Kimi) rate-limited to`:free`\n\nmodels:**1,000 req/day after a one-time $10 top-up**;** 50/day without**.** Catch:**free traffic may be logged/used for upstream training.** Get a key:**[https://openrouter.ai/](https://openrouter.ai/)\n\nNumbers below are list pricing per 1,000,000 tokens, input then output. **Cached input read** = the price for repeated context (huge cost saver for coding agents that resend the same files).\n\n| Provider / Model | Input | Cached read | Output |\n|---|---|---|---|\nAnthropic Claude Sonnet 4.5 |\n$3.00 | $0.30 | $15.00 |\nAnthropic Claude Haiku 4.5 |\n$1.00 | $0.10 | $5.00 |\nOpenAI gpt-5-codex |\n$1.25 | $0.125 | $10.00 |\nGoogle Gemini 2.5 Pro |\n$1.25 (≤200k ctx) | $0.31 | $10.00 |\nGoogle Gemini 2.5 Flash |\n$0.30 | $0.075 | $2.50 |\nGoogle Gemini 2.5 Flash-Lite |\n$0.10 | low | $0.40 |\nZ.ai GLM-4.6 |\n$0.39 | low | $1.74 |\nMiniMax M2 |\n$0.255 | none | $1.00 |\nAlibaba Qwen3-Coder |\n$0.22 | none | $0.90 |\nDeepSeek V4 Flash |\n$0.14 | $0.0028 | $0.28 |\nDeepSeek V4 Pro (list) |\n$1.74 | $0.0145 | $3.48 |\nDeepSeek V4 Pro (promo to 2026-05-31) |\n$0.435 | $0.0036 | $0.87 |\n\n**Sources to verify:**\n\n- Anthropic:\n[https://www.anthropic.com/api](https://www.anthropic.com/api)(or`platform.claude.com/docs/en/about-claude/pricing`\n\n) - OpenAI:\n[https://developers.openai.com/codex/pricing](https://developers.openai.com/codex/pricing) - Gemini:\n[https://ai.google.dev/gemini-api/docs/pricing](https://ai.google.dev/gemini-api/docs/pricing) - DeepSeek:\n[https://api-docs.deepseek.com/quick_start/pricing](https://api-docs.deepseek.com/quick_start/pricing) - MiniMax:\n[https://platform.minimax.io/docs/pricing/overview](https://platform.minimax.io/docs/pricing/overview) - Z.ai / GLM:\n[https://docs.z.ai/guides/overview/pricing](https://docs.z.ai/guides/overview/pricing) - OpenRouter:\n[https://openrouter.ai/](https://openrouter.ai/)\n\n| Plan | $/month | What you get | Works with |\n|---|---|---|---|\nClaude Pro |\n$20 | 5h session cap + weekly cap (~40–80 Sonnet hrs/week) | Claude Code (native) |\nClaude Max 5x / 20x |\n$100 / $200 | 5× / 20× Pro | Claude Code (native) |\nChatGPT Go / Plus / Pro |\n$8 / $20 / $100 | message-based (15–80 msgs / 5h on Plus) | Codex CLI |\nGitHub Copilot Free / Pro / Pro+ |\n$0 / $10 / $39 | usage credits (moving to credits Jun 2026) | Copilot agent |\nCursor Pro |\n$20 | ~$20 credit pool (~225 Sonnet req/mo) | Cursor IDE |\nZ.ai GLM Coding Plan Lite / Pro |\n$10 / $30 | ~120 prompts/5h (Lite); GLM-5.x models | Claude Code, opencode, Cursor |\nMiniMax Coding Starter |\n$10 | ~100 prompts / 5h on M2.1 | Any |\nopencode Go |\n$10 ($5 first month) | $60/mo usage value across 13 open models | opencode (native) |\nopencode Zen |\npay-as-you-go | per-token gateway, mostly $0.30–$3.20/M | opencode (native) |\n\n**Important:** quota caps on flat subs are usually measured in *prompts* (not tokens). If you blow through \"120 prompts / 5h\" with one heavy session you wait 5h — not great for an active coding day. For predictable cost on a budget, token-metered APIs through one gateway usually win.\n\nPer-prompt is tiny; per-day is what matters. Coding-agent turn averages **~5,000 input + 500 output tokens** (lots more if you load a big file). With **60% cache-hit** (typical for repeat-edit sessions):\n\n| Provider / Model | Tokens per $1 (light coding) | Approx prompts |\n|---|---|---|\n| Claude Sonnet 4.5 | ~250k mixed | ~40 prompts |\n| OpenAI gpt-5-codex | ~330k mixed | ~55 prompts |\n| Claude Haiku 4.5 | ~800k mixed | ~130 prompts |\n| Gemini 2.5 Flash | ~1.5M mixed | ~250 prompts |\n| Z.ai GLM-4.6 | ~1.8M mixed | ~300 prompts |\n| MiniMax M2 / Qwen3-Coder | ~3.5M mixed | ~580 prompts |\n| Gemini Flash-Lite | ~7M mixed | ~1,200 prompts |\n| DeepSeek V4 Flash | ~8M mixed | ~1,400 prompts |\n\n*(Indicative only — real cost depends on your prompt size + cache hit rate.)*\n\n**Token ≈ 4 characters** in English; Burmese script ≈ 1 character per token (worse ratio — same text uses more tokens in Burmese).**Input tokens**= your message + the model's context (files, system prompt, history).** Output tokens**= what the model writes back.** Cache read**(Claude/OpenAI/Gemini) = repeated input cached at ~10% price. Huge saver in coding agents that re-send the same files.** A \"long context\"**model (1M tokens) means you can paste a large codebase; you still pay per token.** Tip:**prefer agents that** stream + truncate**old turns instead of re-sending everything.\n\n**Day 1, no card available:****Gemini CLI**(personal account) → free, instant.** Day 1, with a card / Google account, want flexibility:****opencode + AI Studio API key**→ free Flash-Lite tier, scale to paid later.** Premium quality, ok with spend:****Claude Code** via Anthropic API or AWS Bedrock. Budget $5–10/student/mo, more for heavy use.**Cheapest paid bulk:****DeepSeek V4 Flash**(~$0.14/$0.28 per M tokens) via opencode.** Predictable flat:****Z.ai GLM Coding Plan Lite**($10/mo, works in Claude Code) or** opencode Go**($10/mo bundle).\n\nThe bootcamp will publish a shared proxy when it's deployed + secured — at that point students will get a single virtual key that routes to whatever model the proxy is configured for. Until then, pick from the list above.\n\n**Prices verified early-mid 2026.** Confirm on each provider's official pricing page before quoting externally.**Free-tier limits change.** Google dropped 2.5 Pro free tier in April 2026; Flash dropped to ~250/day. OpenRouter's`:free`\n\nrate-limit can change without notice.**Cache assumptions matter.** A 60% cache-hit rate is realistic for coding agents that revisit the same files; one-shot questions get ~0% cache and cost more.**Quotas in prompts vs tokens are different things.** Flat subs cap prompts; APIs cap tokens. A \"120 prompts/5h\" plan is not equivalent to \"X tokens/5h\" — measure your usage profile before committing.**Burmese costs ~2× more in tokens than English** for the same meaning. Plan for it on bilingual projects.\n\nIf you find a current price that disagrees with this table, **the live provider page wins** — open a PR to fix this doc.", "url": "https://wpnews.pro/news/vibe-code-tours-student-setup-linux-macos-wsl-nvm-node22-uv-python3-12-git-code", "canonical_source": "https://gist.github.com/kokoye2007/e8d30985faca6292a50bdd3b0409105f", "published_at": "2026-05-26 12:21:29+00:00", "updated_at": "2026-05-29 15:43:57.527764+00:00", "lang": "en", "topics": ["ai-tools", "ai-products", "large-language-models", "generative-ai", "artificial-intelligence"], "entities": ["Gemini CLI", "Claude Code", "Anthropic", "DeepSeek", "Google", "Amazon Bedrock", "LiteLLM", "opencode"], "alternates": {"html": "https://wpnews.pro/news/vibe-code-tours-student-setup-linux-macos-wsl-nvm-node22-uv-python3-12-git-code", "markdown": "https://wpnews.pro/news/vibe-code-tours-student-setup-linux-macos-wsl-nvm-node22-uv-python3-12-git-code.md", "text": "https://wpnews.pro/news/vibe-code-tours-student-setup-linux-macos-wsl-nvm-node22-uv-python3-12-git-code.txt", "jsonld": "https://wpnews.pro/news/vibe-code-tours-student-setup-linux-macos-wsl-nvm-node22-uv-python3-12-git-code.jsonld"}}