vs
Relace vs Boundless
Relace makes small tool models for apply, search and compaction in coding agents. Boundless serves general open and closed models for the agent itself.
By The Subconscious Team · Updated
Relace vs Boundless: key differences
Relace's relace-apply-3 merges lazy edit snippets into files at about 10,000 tokens per second with up to 128K tokens of input and output, and returns an error past that limit. Its compaction model runs at 50,000 tokens per second, and it adds agentic codebase search and source control. It does no general-purpose model serving, and it can be self-hosted. Boundless serves general models through an OpenAI and Anthropic compatible API, from DeepSeek V4 Flash at $0.04 in and $0.21 out per million tokens to Kimi K3 at $2.30 in and $11.40 out, most with 1M context, plus resold GPT and Claude models.
The two fit together. Boundless can run the model that plans and writes code, and Relace can take apply, search and compaction off that model. Relace's self-hosted option suits teams that keep code in-house, while Boundless keeps zero data retention on its hosted API and offers reserved capacity tuned by its engineers.
What Relace and Boundless do
Relace
Relace trains small, fast models that act as tools for coding agents. Its best-known product is Instant Apply: a frontier model writes a lazy edit snippet, and relace-apply-3 merges it into the original file at about 10,000 tokens per second with 128K tokens of input and output. Relace says this runs over 3x faster and cheaper than having the big model rewrite the file. It exposes both a REST endpoint and an OpenAI compatible one, and the model is also listed on OpenRouter.
Example models: relace-apply-3, Relace agentic search
Full Relace profileBoundless
Boundless launched in September 2026 as an inference partner for AI-native startups. It sells two things. The Model API serves pre-tuned models on shared capacity, billed per token from prepaid credits with a budget per API key. Tailored inference reserves capacity for one customer, and Boundless engineers pick the model on the customer’s own evals, tune precision, batching, caching, speculative decoding and parallelism, then run the deployment against uptime, throughput and latency targets written into the contract. A free two-week evaluation tests those targets on real traffic against the customer’s current provider first.
Example models: GLM-5.3, DeepSeek V4.1 Flash, Kimi K3
Full Boundless profileShould you choose Relace or Boundless?
Relace vs Boundless at a glance
| Attribute | Relace | Boundless |
|---|---|---|
| Model access | Specialist models | Open weights, plus closed GPT and Claude |
| Flagship models | relace-apply-3, agentic search | GLM-5.3, DeepSeek V4.1 Flash, Kimi K3 |
| Speed | ~10,000 tok/s apply | 342 tok/s on DeepSeek V4.1 Flash 1017 (Vercel) |
| Price | 3x+ cheaper than full rewrites | $0.04–$2.30 in, $0.21–$11.40 out per 1M (open) |
| Customization | Unknown | Serving-stack tuning, no fine-tuning |
| Deployment | Hosted API or self-hosted | Serverless API, reserved tailored capacity |
| Long context | 128K max | 1M context on most models |
Frequently asked questions
What is the difference between Relace and Boundless?
Relace makes small tool models for apply, search and compaction in coding agents. Boundless serves general open and closed models for the agent itself.
When should I choose Relace over Boundless?
Instant apply up to 128K tokens; Fast compaction of agent context; Self-hosted deployment.
When should I choose Boundless over Relace?
General models for the agent loop; 1M context on most models; Zero data retention.
Is Relace or Boundless cheaper?
Relace: 3x+ cheaper than full rewrites. Boundless: $0.04–$2.30 in, $0.21–$11.40 out per 1M (open). The cheaper choice depends on the model and workload.
Which has more context, Relace or Boundless?
Relace: 128K max. Boundless: 1M context on most models.
Related comparisons
Running long-horizon agents?
If your agents run past 200K tokens, compare both against Subconscious. Our inference stack treats a long-horizon trace as the primary workload, so speed, cost, and accuracy hold up deep into the trace.