Tabnine

Last verified: Jun 8, 2026

19 sourced changes·95% official·verified against official docs, changelogs & Wayback·last checked Jun 8, 2026

01 / VERDICT

Stable

Low recent constraint volatility — few meaningful restrictions in the last 12 months.

WHY STABLE — trend: stable · drift -2% over 90d · 1 tightening in 90d

Constraint Stability

58/100

higher = fewer recent restrictions

stable-2% / 90d

Tabnine has stayed stable, with minimal changes to its limits over time.

Most noticeable: All free-tier users lost access; solo devs/students have no free path — hard paywall at $39/seat

Recency

days since it last tightened

Velocity

3.7

tightenings per year

Silent

of tightenings were unannounced

Reversals

44%

of tightenings later reversed (same dimension)

Tabnine constraint metrics — computed from 9 tracked tightenings · median 15 days between tightenings. Original to this site; not published by the vendor.

IMPACT BY PERSONA

Solo / occasional

Brief, occasional use is rarely touched by limit changes.

low

Power user / daily heavy

Exposed to usage-limit, rate / frequency tightening.

medium

Team / shared workflow

Exposed to quota, model-availability, usage-based cost changes across the team.

medium

Startup / cost-scaling

Exposed to usage-based cost as usage scales.

medium

Enterprise / long-term dependency

Continuity exposure: model-continuity, ownership / policy shifts.

medium

02 / ANALYSIS

Tabnine has migrated about as plainly as any coding tool here from free and individual-friendly toward enterprise-only pricing. A May 2024 easing added GPT-4o to Pro chat at no extra charge, but the structural moves all tightened. The entry-level Dev plan (~$9–12/seat) was sunset from January 2025, erasing the mid-tier between free and $39/seat; then April 2025 fully sunset the permanent free Basic tier — expanded only months earlier with Claude 3.5 Sonnet and agents — dropping new users onto a hard paywall from $39/seat with no free path for solo devs or students. November 2025 added an Agentic Platform at $59/seat (+51%), with LLM consumption billed pass-through at provider cost +5%, layering unbounded variable cost onto the seat fee. Through mid-2026 the release notes keep narrowing: local workspace indexing removed in favor of the remote Context Engine (a privacy and offline regression), agent input size capped against oversized requests, and a fair-use ceiling of 1B tokens/month introduced where none existed, high enough to bind only extreme volume. Neutral admin controls (per-team seat caps, a YOLO-mode restriction toggle) and small easings (thinking transparency, an Azure OpenAI fix) fill out the period. The repositioning is unmistakable — enterprise seats plus metered agentic consumption, with the free and low-cost paths individuals once used deprecated. Below one $39 seat this is no longer a tool to plan around, and even paid, agentic use carries pass-through LLM cost that scales with how much you run it.

03 / CURRENT LIMITS

Current state + recent changes

Code Assistant ($39/user/mo): AI completions + chat, multi-LLM choice, flexible deployment, zero code retention
Agentic Platform ($59/user/mo): Adds agentic workflows, CLI, Context Engine, MCP; LLM usage billed pass-through +5%

Last verified: Jun 8, 2026

RECENT LIMIT & PRICING CHANGES

BeforeAfter

On-prem deployments were subject to the 1B tokens/month fair-use consumption limit introduced in July 2026.↑Enterprise Capabilities: Removed the 1B monthly token consumption limit for on-prem deployments (v6.4.5, Jul 26, 2026).

Jul 2026·source

Default Tabnine model context limit (and fallback model-profile context window): 180,000 tokens↑Unified the default Tabnine model context limit at 200,000 tokens (Tabnine CLI v0.29.1, shipped with v6.4.3) (+11%)

Jul 2026·source

No stated cap on token consumption.↓Fair-use tiered token limit introduced: 1B tokens/month for regular users.

Jul 2026·source

Subagents ran without default execution limits; local subagents had hardcoded policy bypasses.↓Added default execution limits for subagents, removed the hardcoded policy bypasses for local subagents, and added safeguards against subagent bypass in Plan Mode (Tabnine CLI v0.29.0, documented under v6.4.0).

Jul 2026·source

04 / WHERE IT MATTERS

Which usage patterns the recent changes affect — each line backed by the events driving it.

Agentic / high-volume costmedium exposure

Jul 2026↑Reverses the 1B tokens/month cap for on-prem deployments, ~3 weeks after it was introduced.·source

Jul 2026↓A monthly fair-use ceiling of 1B tokens is newly imposed where none existed; the ceiling is high, so it binds only extreme-volume agentic workflows.·source

Jul 2026↓Subagent runs now hit a default execution cap and can no longer bypass org policy locally or via Plan Mode.·source

Nov 2025↓Agentic capabilities cost +51% plus unbounded variable LLM consumption via pass-through·source

Long-context workloadsmedium exposure

Jul 2026↑Increased context window allows for larger codebases and longer chat sessions.·source

Jun 2026↓Users with very large context inputs may experience truncation or reduced context availability in agent interactions.·source

Jun 2026↓Users relying on local indexing must now use the remote Context Engine, which may impact privacy or offline capabilities.·source

High-frequency usagemedium exposure

Nov 2025↓Agentic capabilities cost +51% plus unbounded variable LLM consumption via pass-through·source

Apr 2025↓All free-tier users lost access; solo devs/students have no free path — hard paywall at $39/seat·source

Jan 2025↓Individual devs/small teams priced out — nothing between (gone) free and $39/seat·source

Sustained sessionslow exposure

Apr 2025↓All free-tier users lost access; solo devs/students have no free path — hard paywall at $39/seat·source

Jan 2025↓Individual devs/small teams priced out — nothing between (gone) free and $39/seat·source

Specific-model dependencelow exposure

Jul 2026↑Restores GPT model access for Agent requests on Enterprise SaaS deployments after a period of intermittent failures.·source

Jul 2026↓Prevents users from bypassing disabled model settings via client-side manipulation, ensuring strict adherence to administrative model access policies.·source

Jun 2026↑Resolves a 404 error preventing access to specific Azure OpenAI models in IDE Chat/Agent.·source

Jun 2026↑Provides visibility into model reasoning for self-managed models, improving debugging and transparency.·source

May 2024↑Pro users gain a faster frontier model; free users remain on the proprietary model only·source

05 / WHAT CHANGED

Technical 6 · Product 9 · Business 1 · Policy 3

Apr 2025
↓ All free-tier users lost access; solo devs/students have no free path — hard paywall at $39/seat
productImpact High · source
Jan 2025
↓ Individual devs/small teams priced out — nothing between (gone) free and $39/seat
productImpact High · source · secondary source
May 2024
↑ Pro users gain a faster frontier model; free users remain on the proprietary model only
productImpact Low · source

▸ View full constraint history (18 events)

Jul 2026
↓ Removes unauthorized bypasses for governance and agent feature controls, enforcing stricter policy compliance.
policyImpact Low · source · secondary source
↔ shipped alongside Restores GPT model access for Agent requests on Enterprise SaaS deployments after a period of intermittent failures. (Jul 2026) — more for more, not a pure cut.
Jul 2026
↑ Restores GPT model access for Agent requests on Enterprise SaaS deployments after a period of intermittent failures.
productImpact Low · source · secondary source
Jul 2026
↑ Reverses the 1B tokens/month cap for on-prem deployments, ~3 weeks after it was introduced.
technicalImpact Low · source · secondary source
Jul 2026
↓ Prevents users from bypassing disabled model settings via client-side manipulation, ensuring strict adherence to administrative model access policies.
productImpact Medium · source · secondary source
↔ shipped alongside Restores GPT model access for Agent requests on Enterprise SaaS deployments after a period of intermittent failures. (Jul 2026) — more for more, not a pure cut.
Jul 2026
→ Optional admin-configurable access control (Azure DevOps only): admins can require an account connection so users only receive context from repositories they are authorized to access. Off by default; enabled per account.
policyImpact Low · source · secondary source
Jul 2026
↑ Increased context window allows for larger codebases and longer chat sessions.
technicalImpact Low · source · secondary source
Jul 2026
↓ A monthly fair-use ceiling of 1B tokens is newly imposed where none existed; the ceiling is high, so it binds only extreme-volume agentic workflows.
technicalImpact Low · source · secondary source
↔ shipped alongside Resolves a 404 error preventing access to specific Azure OpenAI models in IDE Chat/Agent. (Jun 2026) — more for more, not a pure cut.
Jul 2026
→ Adds an opt-in admin governance control; not vendor-imposed on users. Only constrains users whose own org chooses to enable it.
policyImpact Low · source · secondary source
Jul 2026
↓ Subagent runs now hit a default execution cap and can no longer bypass org policy locally or via Plan Mode.
technicalImpact Medium · source · secondary source
↔ shipped alongside Resolves a 404 error preventing access to specific Azure OpenAI models in IDE Chat/Agent. (Jun 2026) — more for more, not a pure cut.
Jun 2026
↑ Resolves a 404 error preventing access to specific Azure OpenAI models in IDE Chat/Agent.
productImpact Low · source · secondary source
Jun 2026
↓ Users with very large context inputs may experience truncation or reduced context availability in agent interactions.
technicalImpact Medium · source · secondary source
↔ shipped alongside Provides visibility into model reasoning for self-managed models, improving debugging and transparency. (Jun 2026) — more for more, not a pure cut.
Jun 2026
→ New admin-controlled capability (opt-in) to enforce seat limits per team; not a vendor-imposed reduction.
technicalImpact Low · source · secondary source
Jun 2026
→ UI relocation — Test/Review tabs replaced by /test and /review slash commands; capability retained, workflow changes.
productImpact Low · source · secondary source
Jun 2026
↓ Users relying on local indexing must now use the remote Context Engine, which may impact privacy or offline capabilities.
productImpact Medium · source · secondary source
↔ shipped alongside Provides visibility into model reasoning for self-managed models, improving debugging and transparency. (Jun 2026) — more for more, not a pure cut.
Nov 2025
↓ Agentic capabilities cost +51% plus unbounded variable LLM consumption via pass-through
businessImpact Medium · source
Apr 2025
↓ All free-tier users lost access; solo devs/students have no free path — hard paywall at $39/seat
productImpact High · source
Jan 2025
↓ Individual devs/small teams priced out — nothing between (gone) free and $39/seat
productImpact High · source · secondary source
May 2024
↑ Pro users gain a faster frontier model; free users remain on the proprietary model only
productImpact Low · source

▸ Other product updates (1) — feature launches, not constraint changes

Jun 2026
↑ Provides visibility into model reasoning for self-managed models, improving debugging and transparency.
productImpact Low · source · secondary source

06 / OUTLOOK

Volatile

Volatile. Last 12 months: 1 pricing, 6 limit/quota, 6 model-availability changes (7 tightenings).

Observed past cadence — not a prediction.

07 / USER SIGNALS

What users report

• “Tabnine discontinued their Dev plan and pivoted to enterprise-only pricing.” source

08 / ALTERNATIVES

ALTERNATIVES

By constraint history — not capability.

More stable:Gemini, Grok, v0

Similar risk:Perplexity, Claude, Devin, Mistral Le Chat

→ Full comparison matrix

09 / RELATED

Tabnine removed its free plan and raised prices →