Duolingo Max

https://blog.duolingo.com/duolingo-max/

Adds GPT-4-powered Roleplay (character conversations, with cross-call memory for "Lily") and Explain My Answer on top of Duolingo's existing spaced-repetition engine, which tracks per-word and per-skill mastery state across all sessions.

At a glance

Type
SRS state + GPT-4 roleplay continuity
Tier
T1
Created
2011 (Duolingo founded 2011; Duolingo Max tier launched 2023)
Latest release
not applicable — not OSS
License
not applicable — not OSS
GitHub
not applicable — no GitHub repo
Pricing
Free + paid
Funding
$183M total raised (CapitalG/NEA/General Atlantic); public company (DUOL)

Taxonomy

storage
relational
retrieval
injection
persistence
lifelong
update
overwrite
unit
profile
governance
inspectable
conflict
overwrite

When to use

Optimised for: learner profile + spaced repetition + assessment-driven adaptation

Anti-fit: not for non-educational use cases

Pros & cons

Pros

Memory grounded in a proven SRS with decades of research behind it; Explain My Answer gives context-specific error feedback rather than generic responses.

Cons

GPT-4 features sit on top of a lesson-completion model — no holistic learner profile; Roleplay memory is character-scoped, not learning-goal-scoped; $168/yr Max pricing.

Claims & capabilities

~9% of 12.2M paid subscribers on Max by end-2025 (~1.1M users). 78% of Roleplay users reported better real-world conversation readiness in internal surveys.

Technical surface

API surface
searched not found
Backend storage
searched not found
Deployment
Managed-only
Embedding model
searched not found
Multi-tenancy
searched not found
MCP
no MCP support advertised — vertical product, no MCP server / client integration documented
A2A
no A2A protocol support advertised — vertical product, no A2A integration documented
OpenTelemetry
no OpenTelemetry integration advertised — vendor logs/observability not publicly documented

Similar systems

Other vertical / domain-specific ai memory in the catalog, ranked by inbound references.

  • NVIDIA ReMEmbR T3

    Builds long-horizon memory by captioning video segments with VILA, storing captions with timestamps + 3D position coordinates in MilvusDB. At query time, LLM iterates retrieval across text, time, and position modalities. Deployed on Nova Carter robot (Jetson Orin).

  • Abridge T1

    Clinician-assist ambient documentation. Source mapping: every AI-generated summary element traced back to the source utterance. Audit-and-trust layer over episodic memory. Built on proprietary 1.5M+ medical-encounter dataset.

  • ASAPP GenerativeAgent T1

    Treats memory as first-class architecture. Captures the digital footprint of every interaction; retrieves preference and history at engagement time. Public example: airline knowing a frequent flyer wants aisle seats with her son — preference-aware, not just history-lookup.

  • BenevolentAI T1

    Target identification / drug repurposing / mechanism tracing. 85+ data sources, petabyte-scale, rebuilt every few weeks. Wet-lab results re-enter the graph and shift downstream predictions — institutional experimental memory closing a feedback loop.

  • Causaly T1

    Drug discovery / target identification / causal mechanism tracing. The graph is the memory: 7 years of curated biomedical cause-effect relationships compounding with each new ingestion. Scientific RAG retrieves from a versioned causal substrate.

  • Character.ai T1

    Chat Memories (user-defined facts), auto-memories for c.ai+ subscribers, pinned memories, in-context retention. PipSqueak 2 model (April 2026) reduces in-conversation drift. Memory Visualization meter forthcoming.

Row last verified 2026-05-14. Catalog data is CC-BY-4.0 — see how to read this.