Back to articles

Data-led article

Low-cost coding-agent API models for a 7k-input workload

Coding-agent cost can move quickly when every task repeats the same repo context across several turns. This comparison derives a current six-route shortlist from the catalog, ranks it by uncached standard-rate cost for one workload, then hands the same workload to the Calculator for cache-sensitive estimates.

Coding-agent workload Data as of catalog generation

Scenario

The workload being priced

The table uses current catalog rows and uncached standard input/output rates. Use the Calculator to test listed cached-input pricing when a route exposes it, and verify final pricing with the provider before production use.

Input / iteration
7,000 tokens

2k prompt plus 5k repo context.

Output / iteration
1,500 tokens

Enough room for a multi-step reply.

Iterations / task
3

Repeated turns on the same task.

Tasks / month
500

A small recurring automation workload.

Cost screen

Low-cost candidates in this workload

Candidates are current chat routes with standard input/output prices, at least 7k context, function calling listed, and no retired lifecycle status. They are sorted by uncached monthly cost; function calling is only an eligibility signal, not proof of coding quality.

Model Context Input / 1M Output / 1M Monthly cost
Qwen2.5-Coder-7B
nebius
32.8K $0.0100 / 1M tokens $0.0300 / 1M tokens $0.17
llama3.2-11b-vision-instruct
lambda_ai
131.1K $0.0150 / 1M tokens $0.0250 / 1M tokens $0.21
llama3.2-3b-instruct
lambda_ai
131.1K $0.0150 / 1M tokens $0.0250 / 1M tokens $0.21
Llama-3.2-3B-Instruct
deepinfra
131.1K $0.0200 / 1M tokens $0.0200 / 1M tokens $0.26
Meta-Llama-3.1-8B-Instruct-Turbo
deepinfra
131.1K $0.0200 / 1M tokens $0.0300 / 1M tokens $0.28
Mistral-Nemo-Instruct-2407
deepinfra
131.1K $0.0200 / 1M tokens $0.0400 / 1M tokens $0.30

The baseline is 7,000 input tokens and 1,500 output tokens across 1,500 monthly turns. The Calculator opens with the same workload and a 0% cache assumption; adjust cache reuse only when your selected route lists cached-input pricing.

Pricing data sourced from our catalog. Check data sources for provenance and freshness.

Caveats

What this comparison does not prove

The site has sparse coding benchmark coverage for this low-cost slice, so this article does not rank models by coding quality. The function-calling flag is only an eligibility filter and does not prove agent quality. Provider discounts, region-specific pricing, rate limits, tool-use charges, and unlisted cache billing are outside this baseline. Cache savings depend on the selected route exposing cached-input pricing. Treat this as a cost shortlist, then test the model on your own coding tasks.