Key Terms
- Gemini Code Assist - Google's AI coding assistant product, available in three tiers: Free (for individuals), Standard (for business), and Enterprise (for business with code customization). Supports VS Code, JetBrains IDEs, Android Studio, and Gemini CLI. Source: Codeassist
- Gemini CLI - an open-source AI agent that runs in the terminal, bringing Gemini models to the command line. Shares daily model request quotas with Gemini Code Assist agent mode. Source: Geminicli
- Agent mode - a multi-step, collaborative reasoning agent within Gemini Code Assist chat that can perform complex software development tasks using system tools and MCP (Model Context Protocol) servers. Source: Google – Use Agentic Chat Pair Programmer
- Code customization - an Enterprise-only feature that augments the model with your private codebases from GitHub, GitLab, or Bitbucket for tailored suggestions. Source: Google – Code Customization Overview
- Google Developer Program - a membership program with Standard (free), Premium ($299/year including Gemini Code Assist Standard), and Enterprise ($75/developer/month including Gemini Code Assist Enterprise) tiers. Source: Google – Program
- IP indemnification - Google covers legal costs if Gemini Code Assist output infringes third-party copyright. Included in Standard and Enterprise tiers. Source: Google – Works
Latest Changes
First report for this supplier. All models, plans, and pricing are listed as current state.
- New model: Gemini 3 and 3.1 Pro launched. Gemini 3.1 Pro Thinking (High): 44.4% HLE, 80.6% SWE-Bench Verified, 68.5% Terminal-Bench 2.0.
- New model: Gemini 3 Flash (Preview) and Gemini 3.1 Flash-Lite (Preview) available.
- Feature added: Gemini CLI (open-source) for terminal-based coding.
- Feature added: Prepay billing option for Gemini API.
- Feature added: Gemini App native macOS client.
- Feature added: Google Antigravity agentic development platform announced.
- Feature added: AI Agents Vibe Coding course from Google and Kaggle.
Plans
Gemini Code Assist for Individuals
| Plan | Price | Usage Limits | Key Inclusions |
|---|---|---|---|
| Free | $0/user/month | 6,000 code-related requests/day, 240 chat requests/day, 1,000 model requests/day (Gemini CLI + Agent mode shared) | Code completion/generation, conversational assistant, Gemini CLI, agent mode, Gemini in Firebase, source citations, IP indemnification, 1M token context window |
Gemini Code Assist for Business (Google Cloud)
| Plan | Price (annual commitment) | Price (monthly commitment) | Usage Limits | Key Inclusions |
|---|---|---|---|---|
| Standard | $19/user/month | $22.80/user/month | 1,500 model requests/day (Gemini CLI + Agent mode shared) | Everything in Free + enterprise-grade security, Gemini in Colab Enterprise, Database Development Assistance, BigQuery Data Insights, Gemini in Cloud Run, admin controls |
| Enterprise | $45/user/month | $54/user/month | 2,000 model requests/day (Gemini CLI + Agent mode shared) | Everything in Standard + code customization, Gemini in Apigee, Gemini in Application Integration, Gemini Cloud Assist |
Alternative: Hourly billing (Google Cloud)
| Plan | Monthly commitment | Annual commitment |
|---|---|---|
| Standard | $0.031/hour | $0.026/hour |
| Enterprise | $0.074/hour | $0.062/hour |
Google Developer Program
| Tier | Price | Includes |
|---|---|---|
| Standard | Free | Foundational services, complementary with Gemini Code Assist for individuals |
| Premium | $299/year | Gemini Code Assist Standard, $1,000+ in Google Cloud credits, hands-on labs, certification voucher |
| Enterprise | $75/developer/month | Gemini Code Assist Enterprise, hands-on labs, risk-managed sandbox |
Source: Codeassist and Google – Pricing
API Pricing
Per-token pricing on Vertex AI for Gemini models. All prices are Standard tier at <=200K context length.
Google Gemini Models
| Model | Input ($/MTok) | Cached Input ($/MTok) | Output ($/MTok) | Notes |
|---|---|---|---|---|
| Gemini 3.1 Pro Preview | $2.00 | $0.20 | $12.00 | >200K: $4.00/$0.40/$18.00 |
| Gemini 3 Pro Preview | $2.00 | $0.20 | $12.00 | >200K: $4.00/$0.40/$18.00 |
| Gemini 3 Flash Preview | $0.50 | $0.05 | $3.00 | Same rate at all context lengths |
| Gemini 3.1 Flash-Lite Preview | $0.25 | $0.03 | $1.50 | Same rate at all context lengths |
| Gemini 2.5 Pro | $1.25 | $0.13 | $10.00 | >200K: $2.50/$0.25/$15.00 |
| Gemini 2.5 Flash | $0.30 | $0.03 | $2.50 | Same rate at all context lengths |
| Gemini 2.5 Flash Lite | $0.10 | $0.01 | $0.40 | Same rate at all context lengths |
Terms explained:
- Cached Input ($/MTok) - price per 1M tokens when using prompt caching, where previously seen input tokens are reused from cache at a reduced rate. Google charges 10% of the standard input price for cached tokens.
- Standard tier - default pay-per-use pricing. Google also offers Priority tier at 1.8x standard rates for lower-latency serving, and Flex/Batch tier at 50% off standard rates for best-effort latency.
- Context Cache Storage - stored context caches incur a separate holding cost of $1.00 to $4.50 per 1M tokens per hour, depending on model and context length.
Source: Google – Pricing
Partner Models on Vertex AI
Google resells third-party frontier models through Vertex AI. Pricing as of April 2026:
| Model | Input ($/MTok) | Output ($/MTok) | Notes |
|---|---|---|---|
| Claude Opus 4.7 | $5.00 | $25.00 | Anthropic's flagship model |
| Grok 4.20 | $2.00 | $6.00 | xAI's frontier model |
| Llama 4 Scout | $0.25 | $0.70 | Meta's lightweight model |
| Mistral Medium 3 | $0.40 | $2.00 | Mistral's mid-tier model |
| Codestral 2 | $0.30 | $0.90 | Mistral's code-specialized model |
Source: Google – Pricing
Model Performance / Benchmarks
| Benchmark | Gemini 3.1 Pro Thinking (High) | Gemini 3 Pro Thinking (High) |
|---|---|---|
| Humanity's Last Exam | 44.4% | - |
| SWE-Bench Verified | 80.6% | - |
| Terminal-Bench 2.0 | 68.5% | - |
| ARC-AGI-2 | - | 77.1% |
| GPQA Diamond | - | 94.3% |
Partner validations:
- GitHub: 35% higher accuracy for Gemini 3 Pro vs Gemini 2.5 Pro in VS Code testing.
- JetBrains: 50% improvement in solved benchmark tasks with Gemini 3 Pro vs Gemini 2.5 Pro.
- Cursor: Co-founders praised Gemini 3 Pro for frontend code quality improvements.
- Cline: "Uses long context far more effectively than Gemini 2.5 Pro and has solved problems that stumped other leading models."
Source: Deepmind – Gemini
Latest News
Gemini Drop April 2026 (April 2026)
Google's monthly Gemini feature update, including new capabilities in the Gemini app. Specific Code Assist features were not detailed in the accessible blog content.
Source: Google – Gemini Drop April 2026
Gemini App Now on Mac (April 2026)
Google released a native macOS app for Gemini, expanding desktop access beyond the browser.
Source: Google – Gemini App Now On Mac Os
Gemini 3 and 3.1 Pro Launch (Recent)
Google launched Gemini 3 as its most intelligent model, with state-of-the-art performance across reasoning, coding, and multimodal benchmarks. Gemini 3.1 Pro adds further improvements. Key benchmarks from Google DeepMind:
- Gemini 3.1 Pro Thinking (High): 44.4% on Humanity's Last Exam, 80.6% on SWE-Bench Verified, 68.5% on Terminal-Bench 2.0
- Gemini 3 Pro Thinking (High): 77.1% on ARC-AGI-2, 94.3% on GPQA Diamond
- Partners integrating Gemini 3 include GitHub Copilot (35% higher accuracy vs Gemini 2.5 Pro), Cursor, Cline, JetBrains (50% improvement over Gemini 2.5 Pro in benchmark tasks), and Figma
Source: Deepmind – Gemini
AI Agents Vibe Coding Course from Google and Kaggle (April 2026)
New free course teaching vibe coding with AI agents, available through Kaggle.
Source: Google – Kaggle Genai Intensive Course Vibe Coding June 2026
Prepay for Gemini API (April 2026)
Google introduced prepaid billing for the Gemini API, giving developers more control over spend with upfront commitments.
Source: Google – Prepay Gemini Api
Google Antigravity (Recent)
Google announced Antigravity, a new agentic development platform described as evolving the IDE into the "agent-first era."
Source: Antigravity
Community Signals
Gemini 3 Pro Reception
Gemini 3 Pro's benchmark performance generated significant discussion on HackerNews, with multiple commenters noting the strong SWE-Bench Verified score (80.6%) and competitive Terminal-Bench 2.0 result (68.5%). Cursor's co-founders praised the model for frontend code quality improvements. Cline's Head of AI reported that Gemini 3 Pro "uses long context far more effectively than Gemini 2.5 Pro and has solved problems that stumped other leading models." JetBrains reported 50% improvement in solved benchmark tasks with Gemini 3 Pro vs Gemini 2.5 Pro. GitHub reported 35% higher accuracy for Gemini 3 Pro vs Gemini 2.5 Pro in VS Code testing.
Free Tier Value
Community on Reddit notes that Gemini Code Assist's free tier is one of the most generous available: 6,000 code requests/day and 240 chat requests/day at no cost. The Gemini CLI (open source) has been well-received as a competitive alternative to Claude Code and Cerebras Code for terminal-based coding.
Source: Deepmind – Gemini
Enterprise Readiness
| Feature | Available? | Details |
|---|---|---|
| SSO (SAML) | Yes | Via Google Cloud Identity on Standard and Enterprise plans. Source: Google – Works |
| SSO (OIDC) | Yes | Via Google Cloud Identity on Standard and Enterprise plans. |
| SCIM | Undisclosed | Not explicitly mentioned for Code Assist. Available in broader Google Cloud Workspace. |
| Audit logs | Yes | Available through Google Cloud Audit Logs. Source: Google – Works |
| IP indemnity | Yes | Included in Standard and Enterprise tiers. Source: Google – Works |
| Data residency | Yes | Via Google Cloud regions. Source: Google – Works |
| HIPAA | Yes | Google Cloud offers BAA for covered services. Source: Google – Hipaa |
| Air-gapped / on-prem | No | Not available for Gemini Code Assist. Google Cloud Vertex AI is cloud-only. |
| SLA | Yes | Google Cloud SLAs apply. Source: Google – Works |
| Admin controls (RBAC) | Yes | Admin controls on Standard and Enterprise plans. Google Cloud IAM for fine-grained access. Source: Codeassist |
Transparency Gaps
| Metric | Status | Notes |
|---|---|---|
| Per-token API pricing | disclosed | Vertex AI publishes input/cached input/output per MTok for all Gemini models. Source: Google – Pricing |
| Code Assist usage limit units | partially disclosed | Requests/day are published, but "code-related requests" vs "chat requests" distinction is vague |
| Model request definitions | partially disclosed | "Model requests" for CLI and Agent mode are shared but definition of a single "request" is not specified |
| Gemini 3 Flash context window | undisclosed | Not published on DeepMind or Code Assist pages |
| Gemini 3.1 Pro context window | undisclosed | Not published on DeepMind or Code Assist pages |
| Context cache storage pricing range | partially disclosed | Published as $1.00 to $4.50 per MTok per hour, but the exact per-model breakdown is not specified |
| Code customization pricing | bundled | Included in Enterprise tier only, no standalone pricing |
| Gemini Cloud Assist paid features | preview pricing | Listed as "no additional charge during preview" but future pricing unknown |
| Hourly billing conversion | disclosed | Monthly/annual commitment prices are clearly listed |