#Models

GPT-GOB ships four models in the GOB-5.5 family. They share the same Cave-Neural Framework architecture but differ in size, depth, and routing strategy.

##The lineup

Model ID	Params	Context	Best for	Latency (p50)
`gob-5.5-scout`	8B	32k	High-volume chat, classification, routing	~180ms ttft
`gob-5.5`	70B	128k	General purpose. The default.	~320ms ttft
`gob-5.5-deep`	70B + DCMx7	128k	Hard reasoning, research, code review	~780ms ttft
`gob-5.5-horde`	405B MoE	256k	Maximum quality, agentic workflows	~520ms ttft

ttft = time to first token. Streaming latency, p50, measured against `api.gpt-gob.ai` from us-east-1.

##Choosing a model

###`gob-5.5-scout` — fast and dirty

8B parameters. Cheap. Use it when you don't need deep reasoning: classification, intent detection, summarization, simple Q&A. Smart enough to know when to escalate to a bigger model.

###`gob-5.5` — the default

70B parameters, 128k context, full Cave-Neural Framework. This is what you want unless you have a reason to pick something else. Hits the sweet spot of price, latency, and quality.

###`gob-5.5-deep` — for the hard ones

Same 70B base as gob-5.5 but runs Deep Context Mining at maximum depth (dcm_passes=7) on every request. Slower and pricier, but finds non-obvious solutions other models miss. Use it for:

▸Multi-step math
▸Code review and architecture decisions
▸Research synthesis
▸Anything where gob-5.5 gives you "kind of right" answers

###`gob-5.5-horde` — flagship

405B parameters in a Mixture-of-Experts setup. Horde Routing dynamically assembles parameter clusters per request, so per-token latency is closer to a 70B model than the params suggest. Top of the leaderboard for Lateral Thinking and Subtext Detection. Most expensive.

##Programmatic listing

bash

curl https://api.gpt-gob.ai/v1/models \
  -H "Authorization: Bearer $GOB_API_KEY"

Returns:

json

{
  "object": "list",
  "data": [
    {
      "id": "gob-5.5-scout",
      "object": "model",
      "created": 1746998400,
      "owned_by": "gpt-gob",
      "context_window": 32768,
      "max_output_tokens": 8192,
      "supports_streaming": true,
      "supports_tools": true,
      "supports_json_mode": true,
      "params": "8B"
    },
    {
      "id": "gob-5.5",
      "object": "model",
      "created": 1746998400,
      "owned_by": "gpt-gob",
      "context_window": 131072,
      "max_output_tokens": 16384,
      "supports_streaming": true,
      "supports_tools": true,
      "supports_json_mode": true,
      "params": "70B"
    },
    {
      "id": "gob-5.5-deep",
      "object": "model",
      "created": 1746998400,
      "owned_by": "gpt-gob",
      "context_window": 131072,
      "max_output_tokens": 16384,
      "supports_streaming": true,
      "supports_tools": true,
      "supports_json_mode": true,
      "params": "70B+dcmx7"
    },
    {
      "id": "gob-5.5-horde",
      "object": "model",
      "created": 1746998400,
      "owned_by": "gpt-gob",
      "context_window": 262144,
      "max_output_tokens": 32768,
      "supports_streaming": true,
      "supports_tools": true,
      "supports_json_mode": true,
      "params": "405B-moe"
    }
  ]
}

##Get a single model

bash

curl https://api.gpt-gob.ai/v1/models/gob-5.5 \
  -H "Authorization: Bearer $GOB_API_KEY"

##Aliases

For convenience, version-less aliases always point to the latest stable release in each tier:

Alias	Currently resolves to
`gob-scout`	`gob-5.5-scout`
`gob`	`gob-5.5`
`gob-deep`	`gob-5.5-deep`
`gob-horde`	`gob-5.5-horde`
`gob-latest`	`gob-5.5-horde`

Aliases will track new releases automatically. Pin to a versioned model ID in production if you need behavior to be stable.

##Deprecation policy

We give 6 months notice before any model is removed from the API. Deprecated models stay accessible at a reduced rate limit until removal. See Changelog for the current schedule.