Beyond Chinchilla-Optimal: Accounting for Inference in Language Model Scaling Laws
read the original abstract
Large language model (LLM) scaling laws are empirical formulas that estimate changes in model quality as a result of increasing parameter count and training data. However, these formulas, including the popular Deepmind Chinchilla scaling laws, neglect to include the cost of inference. We modify the Chinchilla scaling laws to calculate the optimal LLM parameter count and pre-training data size to train and deploy a model of a given quality and inference demand. We conduct our analysis both in terms of a compute budget and real-world costs and find that LLM researchers expecting reasonably large inference demand (~1B requests) should train models smaller and longer than Chinchilla-optimal. Furthermore, we train 47 models of varying sizes and parameter counts to validate our formula and find that model quality continues to improve as we scale tokens per parameter to extreme ranges (up to 10,000). Finally, we ablate the procedure used to fit the Chinchilla scaling law coefficients and find that developing scaling laws only from data collected at typical token/parameter ratios overestimates the impact of additional tokens at these extreme ranges.
This paper has not been read by Pith yet.
Forward citations
Cited by 16 Pith papers
-
The Inference-Compute Frontier and a Latency-Efficient Architecture for Limit Order Book Prediction
Empirical power-law frontier between predictive loss and structural forward work in LOB models extrapolates to held-out high-compute architectures with R²=0.941, motivating FastBiNLOB which exceeds SOTA macro-F1 at lo...
-
Neural Garbage Collection: Learning to Forget while Learning to Reason
Language models learn to evict KV cache entries end-to-end via reinforcement learning from outcome reward alone, achieving 2-3x cache compression while maintaining accuracy on Countdown, AMC, and AIME tasks.
-
BoostTaxo: Zero-Shot Taxonomy Induction via Boosting-Style Agentic Reasoning and Constraint-Aware Calibration
BoostTaxo introduces a boosting-style LLM framework for zero-shot taxonomy induction that uses hybrid candidate selection and constraint-aware calibration to achieve superior or comparable performance to prior methods...
-
Memory as a Wasting Asset: Pricing Flash Endurance for Embodied Agents, and the Limits of Doing So
Flash endurance is priced via shadow price η making placement cost-optimal for any sign of value-write correlation χ, with χ positive only in recurrent long-horizon manipulation and the budget binding only on low-endu...
-
Predictable Scaling Laws of Optimal Hyperparameters for LLM Continued Pre-training
Optimal hyperparameters for LLM continued pre-training follow predictable scaling laws derived from proxy models, enabling a two-stage framework that predicts settings from compute budget and checkpoint state to reduc...
-
Policy-Guided Stepwise Model Routing for Cost-Effective Reasoning
A small RL-trained policy for stepwise model routing between LLM sizes improves the accuracy-cost tradeoff on math benchmarks over handcrafted strategies and matches large process reward model methods.
-
When Less is Enough: Efficient Inference via Collaborative Reasoning
A large model generates a compact reasoning signal that a small model uses to solve tasks, reducing the large model's output tokens by up to 60% on benchmarks like AIME and GPQA.
-
Scaling Laws Meet Model Architecture: Toward Inference-Efficient LLMs
A conditional scaling law fitted on over 200 models from 80M to 3B parameters identifies architectures that deliver up to 2.1% higher accuracy and 42% higher inference throughput than LLaMA-3.2 under the same training budget.
-
The Unreasonable Effectiveness of Entropy Minimization in LLM Reasoning
Entropy minimization on self-generated outputs elicits strong reasoning in pretrained LLMs, matching or exceeding supervised RL methods on benchmarks.
-
MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies
MiniCPM 1.2B and 2.4B models reach parity with 7B-13B LLMs via model wind-tunnel scaling and a WSD scheduler that yields a higher optimal data-to-model ratio than Chinchilla scaling.
-
The Thermodynamic Costs of Simple Linear Regression
Thermodynamic lower bounds are approximated for exact and SGD linear regression, producing energy-aware scaling laws for optimal training dataset size given a target generalization error.
-
Trust, but Verify: Peeling Low-Bit Transformer Networks for Training Monitoring
A layer-wise peeling framework creates reference bounds to diagnose under-optimized layers in trained decoder-only transformers, including low-bit and quantized versions.
-
L20-Edu-135M: An Auditable Single-GPU Study of Data-Efficient Small Language Modeling
Trains L20-Edu-135M (134.5M params) on 13B tokens on one GPU, reports 0.415 mean zero-shot score (87% of SmolLM-135M) with full pipeline documentation and notes RLVR reduced GSM8K accuracy.
-
Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale
Technical report announcing Ling-2.6 and Ring-2.6 models with hybrid linear attention, evolutionary CoT, and KPop RL for efficient agentic intelligence at scale.
-
Yi: Open Foundation Models by 01.AI
Yi models are 6B and 34B open foundation models pretrained on 3.1T curated tokens that achieve strong benchmark results through data quality and targeted extensions like long context and vision alignment.
-
Small Language Models (SLMs) Can Still Pack a Punch: A survey (updated 2026)
A literature survey of Small Language Models (1-8B parameters) that can perform comparably or better than larger models, covering general-purpose and task-specific approaches plus creation techniques.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.