by Anysphere (Cursor)In-IDE ExtensionfreemiumRank #16 on Leaderboard

Supermaven

The fastest code autocomplete engine with an unprecedented 1,000,000 token context window.

Overall Index
72.0/ 100
Artificial Analysis benchmark composite
SWE-bench Verified
42.0%
Real-world GitHub issue resolution
DeepSWE v1.1
61.0%
Multi-file software engineering tasks
Terminal-Bench
—
Shell execution & command accuracy
Avg Cost / Task
$0.20
Inference token expenditure per task

Overview & Analysis

Created by Jacob Jackson (creator of Tabnine) and acquired by Cursor/Anysphere, Supermaven is engineered for sheer autocomplete velocity. Its custom Babble neural network boasts an astonishing 1,000,000 token context window that remembers your entire workspace and returns completions in under 100 milliseconds.

Pricing Model: Free tier with unlimited standard completions; Pro at $10/month with 1M token context Babble model.

Free Tier Details: Free forever with unlimited autocomplete using standard model.

Technical Specs & Models

Primary EngineBabble (Proprietary 1M Context Model)
EnvironmentIDE Extension (VS Code, JetBrains, Neovim)
PlatformsmacOS, Linux, Windows
LicenseProprietary
Custom API Keys (BYOK)Bundled Only (No)
Supported Foundation Models:
Babble 1M Context EngineSupermaven Standard

Key Capabilities & Features

Specialized engineering mechanisms implemented in Supermaven.

1,000,000 Token Context Window

Maintains full awareness of recently viewed files, git histories, and cross-file definitions in active memory.

Sub-100ms Inference Latency

Completions appear almost instantaneously as you type, eliminating awkward pauses.

Multi-Editor Extension Ecosystem

Lightweight extensions available for VS Code, JetBrains, and Neovim.

Pros & Cons Evaluation

Forensic analysis of developer experience, cost efficiency, and trade-offs.

Strengths & Advantages
  • Fastest autocomplete response latency in the software engineering industry
  • Massive 1M token context understands deep codebase nuances without vector embeddings
  • Very affordable Pro plan at only $10/month (half the cost of Cursor Pro)
Limitations & Trade-Offs
  • Focused strictly on code autocomplete; does not provide an autonomous multi-file terminal agent
  • No standalone conversational chat interface

Pricing Plans & Tiers

Subscription plans, credit quotas, and commercial deployment options.

Free

$0

For individual developers

  • Unlimited code completions
  • Standard speed model
  • VS Code, JetBrains, Neovim support
Choose Free
Most Popular

Pro

$10/ month

For professional software developers

  • 1,000,000 token context Babble model
  • Sub-100ms response time
  • Higher completion accuracy
  • Priority routing
Choose Pro

Verified Benchmark Evaluations

Standardized evaluation scores recorded on SWE-bench and synthetic code harnesses.

Benchmark HarnessScore / AccuracyEvaluation MetricVerification StatusEvaluation Notes
DeepSWE v1.161.0%Autocomplete AccuracyVerifiedStandard benchmark instance run.

Alternative AI Coding Agents & Comparisons

Compare Supermaven against other top-ranked software engineering assistants.