Glowing Spoon

Industry / Models

Frontier intelligence. Open possibilities.

We work with frontier and open-source models. We select across hosted APIs and open-weight releases based on coding quality, reasoning, multimodal needs, deployment control, and cost per successful task.

01 / AnthropicFrontier API

Claude

Long-running coding, document understanding, and tool-driven workflows with explicit reasoning controls.

  • Agentic coding
  • Visual reasoning
  • Knowledge work
Capabilities & benchmarks
02 / OpenAIFrontier API / coding agent

Codex

An engineering agent backed by OpenAI models for repository work, tool execution, and reviewable software changes.

  • Repository engineering
  • Tool execution
  • Code review
Capabilities & benchmarks
03 / Google DeepMindFrontier API

Gemini

Multimodal intelligence for engineering, long-context reasoning, and workflows that combine visual and textual evidence.

  • Multimodal analysis
  • Long-context reasoning
  • Agentic coding
Capabilities & benchmarks
04 / Qwen / AlibabaAPI + open weights

Qwen

A family spanning hosted reasoning models and open-weight coding checkpoints for configurable deployment.

  • Agentic coding
  • Instruction following
  • Deployment control
Capabilities & benchmarks
05 / DeepSeekAPI + open weights

DeepSeek

Reasoning and agentic task execution with hosted access and an MIT-licensed checkpoint for deployment control.

  • Code agents
  • Reasoning
  • Open-weight deployment
Capabilities & benchmarks
06 / MetaOpen weights

Meta / Llama

An open-weight multimodal MoE model for teams that need control over inference, adaptation, and deployment.

  • Multimodal understanding
  • Custom deployment
  • Multilingual applications
Capabilities & benchmarks
07 / Grok / x.aiFrontier API

Grok

Coding and knowledge-work models for tool-driven engineering and operational assistants.

  • Coding agents
  • Knowledge work
  • Tool-driven workflows
Capabilities & benchmarks

How we choose

Benchmarks inform. Your workload decides.

Each profile names the exact checkpoint, evaluation setup, and source. Scores from different benchmark versions, tool harnesses, reasoning budgets, or release dates are not a single leaderboard. We validate candidates on representative tasks before selecting a production configuration.

Open-source and open-weight are distinct: Qwen’s selected Coder checkpoint is Apache 2.0, DeepSeek’s listed checkpoint is MIT, and Llama uses its own community license. Access and terms are shown on each profile.

Sources checked / 07 November 2026