Startup profile · funding coverage
Remanence funding, valuation and investors
Enterprise AI post-training platform that turns agent execution traces into evaluation environments and specialized customer-owned open-weight models.
Verified company research · funding, investors and latest developments
Keep track of Remanence
Save this profile to your VCT watchlist for a quick return.
Funding, valuation & investors
Answer-first snapshotLatest funding
Seed
$30M · November 17, 2026
Latest known valuation
Not publicly disclosed
Total disclosed equity funding
$30M
Excludes debt, grants, acquisitions, secondaries, and IPO proceeds.
Current status
private
seed · Paris, France
Investors in latest funding
Lead: EQT Ventures , FirstMark Capital
Other: Zero Prime , Emblem , Drysdale , Global Founders Capital , Yellow
Overview
Founded in 2026 by former H Company researchers Pierre-Louis Cedoz (CEO), Hamza Benchekroun (CTO) and Axel Moyal (CPO). Remanence connects agent observability data, identifies failure patterns, synthesizes sandbox tasks and verifiers, post-trains open-weight models by long-horizon reinforcement learning, and offers managed or customer-cloud inference.
EQT Ventures and FirstMark co-led $30M Seed announced Nov 17, with Zero Prime, Emblem, Drysdale, Global Founders Capital and Yellow participating. Customers can own resulting weights and keep data in their infrastructure, per company claims; commercial pricing, independent ROI and reference customers remain undisclosed.
Sources: https://firstmark.com/story/intelligence-that-compounds-why-we-co-led-remanences-30m-seed/ ; https://stories.eqtventures.com/articles/own-don%E2%80%99t-rent-ai-why-we-co-led-remanence%E2%80%99s-30m-seed-round ; https://www.remanencelabs.ai/
Why Remanence is interesting
$30M Seed co-led by EQT Ventures and FirstMark on Nov 17, 2026 backs a shift from renting generic model inference toward improving owned models using failures and corrections from production agents. The company has not disclosed audited customer performance, revenue, valuation or pricing.
Product & use cases
Illustrative technical buyer workflow based on vendor documentation, not hands-on testing: connect agent traces and observability; identify recurring failure modes; construct sandbox tasks and verifiers; post-train selected open-weight model via reinforcement learning; benchmark accuracy, latency and full training/inference cost against a held-out baseline; deploy managed or in customer VPC/on-prem and repeat as traces accumulate. No named SDK integrations, self-service price or independently reproducible benchmarks published.
- Customer-support or legal AI operators exploring dedicated models trained from workflow traces; reported design-partner sectors, not independently verified deployments.
- Enterprise model-serving team evaluating total cost of model ownership versus recurring frontier-model API bills.
- Regulated enterprise exploring VPC/on-prem processing subject to diligence of actual isolation, security and data terms.
Key facts
- $30M Seed announced 17 Nov 2026, co-led by EQT Ventures and FirstMark.
- Vendor says customers own specialized model weights; can deploy in customer infrastructure.
- Pricing, valuation, independently measured gains and named customer deployments not disclosed.
Funding history (newest first)
Seed
2026-11-17 $30M- EQT Ventures (lead)
- FirstMark Capital (lead)
- Zero Prime (participant)
- Emblem (participant)
- Drysdale (participant)
- Global Founders Capital (participant)
- Yellow (participant)
Source: Read source · firstmark.com
Investors in our directory
Funds linked from Remanence's profile — open a fund page for stage focus and related deal articles.
Competitive landscape
Edge: Synthesizes RL environments from task history, aiming to make enterprise-specific post-training less laborious. Weight ownership and infrastructure control may differentiate from hosted proprietary model APIs, but overall quality, retraining overhead and savings are unproven.
Compare end-to-end total cost including trace curation, privacy review, synthetic environment validation, GPU training, serving, evaluation and recurring retraining. Company performance and cost claims lack named audited baselines.
-
Predibase direct
Managed model fine-tuning and serving; compare environment generation and training controls.
-
OpenAI fine-tuning/API agents alternative
Quick managed adoption, but proprietary hosted weights and pricing are a different tradeoff.
-
Hugging Face TRL / verl alternative
Open-source RL tooling requires in-house eval, infrastructure and environment development.
Market / IPO context
Editorial / static context — not a live quote.
Last updated:
Editorial note: AI tools assisted with research, structure, or drafting. Venture Capital Tracker retains human editorial responsibility for factual accuracy, relevance, and source quality before publication.