Models shift to continuous/iterative training

Author Johannes Haus
Last updated
Confirmed · Model Capability · 90% confidence
Predicted: Late 2025 · Updated: 2026-07-20 · Source: ai-2027.com, Late 2025: The World's Most Expensive AI
By this point 'finishes training' is a bit of a misnomer; models are frequently updated to newer versions trained on additional data or partially re-trained to patch some weaknesses.

At a glance

  • Assessment: Confirmed
  • Confidence: 90%
  • Predicted timing: Late 2025
  • Primary source: ai-2027.com, Late 2025: The World's Most Expensive AI

What AI 2027 Predicted

The scenario describes a shift away from discrete model releases toward continuous, iterative training. Rather than training a model from scratch and releasing it as a finished product, labs would frequently update models — training on additional data, patching weaknesses, and releasing incremental versions. The term “finishes training” becomes a misnomer.

How We Track This

We monitor:

  • Frequency of model version updates from major labs (e.g., GPT-4o → 5 → 5.1 → 5.2 → 5.4)
  • Lab announcements describing iterative or continuous training paradigms
  • Versioning patterns suggesting ongoing refinement rather than fresh training runs
  • API model deprecation schedules showing the pace of updates

Current Evidence

The evidence for this prediction is strong. OpenAI’s GPT-5 family demonstrates this pattern clearly: GPT-5 launched in mid-2025, followed by GPT-5.1, GPT-5.2, GPT-5.2-Codex, and GPT-5.4 in rapid succession — each appearing to build on the same base model with iterative improvements. The AI Futures Project’s own grading assessed this as “correct,” noting that “GPT-4o → GPT-5 → GPT-5.1 appear to be continuations of same base model.”

Anthropic has followed a similar pattern with Claude model refreshes (Claude 3.5 Sonnet → Claude 3.7 Sonnet → Claude Opus 4 → Claude Opus 4.5 → Claude Opus 4.6), with intermediate updates and capability patches. Google has also adopted rapid iteration with Gemini model versions.

The pattern appears broadly adopted across the industry: model versioning now resembles software release cycles more than discrete research publications.

NVIDIA describes continuous post-training as a production feedback loop in which deployment surfaces new problems, repeated runs generate rollouts, rewards are verified, and updated weights flow back into training. NVIDIA also identifies Prime Intellect as continuously post-training frontier open models on Blackwell infrastructure. This provides direct vendor evidence for iterative weight updates after deployment, while leaving continual pretraining and daily frontier-model online learning unproven.

Sources:

Counterevidence & Limitations

  • It’s unclear whether iterative updates involve actual continued pretraining on the base model or are primarily post-training refinements (fine-tuning, RLHF, prompt engineering)
  • The distinction between “continuous training” (as AI 2027 envisions — weight updates from new data) and “continuous deployment” (same base model, improved scaffolding) matters for the deeper claim
  • Some version bumps may be more about marketing than fundamental model changes

What Would Change Our Assessment

  • Already confirmed. Would strengthen further if labs explicitly describe online/continual pretraining paradigms
  • Downgrade risk: If evidence emerges that version updates are purely superficial post-training changes with no base model modification (unlikely given observed capability jumps)

Update History

DateUpdate
2026-07-20NVIDIA described continuous post-training as a production feedback loop with repeated rollouts and updated weights, and identified Prime Intellect as continuously post-training frontier open models. This supports iterative training while leaving continual pretraining and daily online learning unproven. Confidence remains 0.90.
2026-03-13Continuous/iterative training now standard practice at frontier labs. AI Futures Project graded this prediction as correct.
2025-12GPT-5.2 released — sixth major GPT version in ~8 months. “Code Red” response to Gemini 3 makes explicit that labs iterate continuously in response to competitive signals. GPT-5.x series demonstrates rapid iterative updates, with multiple versions released within months. Claude model cadence follows similar pattern.
2025-11GPT-5.1 (Nov 12), GPT-5.1-Codex-Max (Nov 18), Gemini 3 six days later, Claude Opus 4.5 six days after that. Three major updates in under a month. The concept of distinct “model releases” is blurring into continuous iteration.
2025-05Claude Opus 4 and Sonnet 4 released together (May 22). Google Gemini 2.5 Flash at I/O (May 20). Three frontier labs shipping major updates within days of each other.
2025-04OpenAI releases GPT-4.1, GPT-4.1-mini, and GPT-4.1-nano simultaneously (April 14) while deprecating GPT-4.5 Preview. Multiple model variants at different capability/cost points shipped together signals iterative, continuous development replacing discrete “generation” releases.