Model catalog

Models ready for agentic tools.

Use these model IDs with any OpenAI-compatible SDK or coding agent pointed at the Aegis proxy. The catalog below mirrors the models NeuralWatt currently exposes, including context limits, capabilities, and token pricing.

Models

Energy rate

$5.00/kWh

API shape

OpenAI compatible

Available Models

Model IDs are the exact strings to place in the `model` field of chat completion requests.

Reference energy uses 0.35 J per input token

Qwen

Qwen3.5 397B

Reasoning

Qwen3.5 397B-A17B — quant-agnostic canonical name for the Qwen3.5 397B family.

Context

262.13K

Input

$0.6900/M

Energy

~1.75 Wh

ReasoningToolsJSON

Output

$4.14/M

Cached input

$0.1725/M

Images

None

Model ID

qwen3.5-397b

Reference request cost$0.0088

Based on a 18K-token reference input before Aegis optimization.

Qwen

Qwen3.5 397B Fast

Fast

Qwen3.5 397B with thinking mode disabled for faster, lower-latency responses. Ideal for straightforward tasks where extended reasoning is unnecessary.

Context

262.13K

Input

$0.6900/M

Energy

~1.17 Wh

ToolsJSON

Output

$4.14/M

Cached input

$0.1725/M

Images

None

Model ID

qwen3.5-397b-fast

Reference request cost$0.0058

Based on a 12K-token reference input before Aegis optimization.

MoonshotAI

Kimi K2.6

Reasoning

Kimi K2.6 — quant-agnostic canonical name for the Kimi K2.6 family.

Context

262.13K

Input

$0.6900/M

Energy

~1.75 Wh

ReasoningVisionToolsJSON

Output

$3.22/M

Cached input

$0.1725/M

Images

Model ID

kimi-k2.6

Reference request cost$0.0088

Based on a 18K-token reference input before Aegis optimization.

MoonshotAI

Kimi K2.6 Fast

Fast

Moonshot Kimi K2.6 with thinking mode disabled for instant, lower-latency responses.

Context

262.13K

Input

$0.6900/M

Energy

~1.17 Wh

VisionToolsJSON

Output

$3.22/M

Cached input

$0.1725/M

Images

Model ID

kimi-k2.6-fast

Reference request cost$0.0058

Based on a 12K-token reference input before Aegis optimization.

Qwen

Qwen3.6 35B

Reasoning

Qwen3.6 35B-A3B — quant-agnostic canonical name for the Qwen3.6 35B family.

Context

131.06K

Input

$0.2900/M

Energy

~1.56 Wh

ReasoningVisionToolsJSON

Output

$1.15/M

Cached input

$0.0725/M

Images

Model ID

qwen3.6-35b

Reference request cost$0.0078

Based on a 16K-token reference input before Aegis optimization.

Qwen

Qwen3.6 35B Fast

Fast

Qwen3.6 35B with thinking mode disabled for instant responses. A lightweight, fast model for rapid iteration.

Context

131.06K

Input

$0.2900/M

Energy

~0.97 Wh

VisionToolsJSON

Output

$1.15/M

Cached input

$0.0725/M

Images

Model ID

qwen3.6-35b-fast

Reference request cost$0.0049

Based on a 10K-token reference input before Aegis optimization.

MoonshotAI

Kimi K2.7 Code

Reasoning

Kimi K2.7 Code — quant-agnostic canonical name for the Kimi K2.7 Code family.

Context

262.13K

Input

$0.9500/M

Energy

~1.75 Wh

ReasoningVisionToolsJSON

Output

$4.00/M

Cached input

$0.2375/M

Images

Model ID

kimi-k2.7-code

Reference request cost$0.0088

Based on a 18K-token reference input before Aegis optimization.

ZhipuAI

GLM-5.2

Reasoning

Private GLM-5.2 test canary

Context

1.05M

Input

$1.45/M

Energy

~1.75 Wh

ReasoningTools

Output

$4.50/M

Cached input

$0.3625/M

Images

None

Model ID

glm-5.2

Reference request cost$0.0088

Based on a 18K-token reference input before Aegis optimization.

ZhipuAI

GLM-5.2 (fast)

Fast

GLM-5.2 with thinking skipped (reasoning_effort=none). Non-thinking tier, mirrors glm-5.1-fast.

Context

1.05M

Input

$1.45/M

Energy

~1.17 Wh

Tools

Output

$4.50/M

Cached input

$0.3625/M

Images

None

Model ID

glm-5.2-fast

Reference request cost$0.0058

Based on a 12K-token reference input before Aegis optimization.

ZhipuAI

GLM-5.2 (short)

Reasoning

GLM-5.2 with a 200K context window and a bounded reasoning budget — faster, lower-energy serving for everyday coding, agentic, and chat workloads under 200K tokens. Private preview (grant-gated).

Context

199.98K

Input

$1.45/M

Energy

~1.36 Wh

ReasoningTools

Output

$4.50/M

Cached input

$0.3625/M

Images

None

Model ID

glm-5.2-short

Reference request cost$0.0068

Based on a 14K-token reference input before Aegis optimization.

ZhipuAI

GLM-5.2 (short, fast)

Fast

GLM-5.2 short with reasoning OFF — the fastest, lowest-energy variant of the 200K pool, for latency-sensitive coding, chat, and tool-use that does not need chain-of-thought. Private preview (grant-gated).

Context

199.98K

Input

$1.45/M

Energy

~0.97 Wh

Tools

Output

$4.50/M

Cached input

$0.3625/M

Images

None

Model ID

glm-5.2-short-fast

Reference request cost$0.0049

Based on a 10K-token reference input before Aegis optimization.

Integration

Paste the ID into your agent config.

Aegis keeps the endpoint OpenAI-compatible. Keep your Aegis key in the API key field, set the base URL once, then choose a model ID from the table for each agent profile.

Model ID	Provider	Context	Input price	Reference energy
qwen3.5-397b	Qwen	262.13K	$0.6900/M	~1.75 Wh
qwen3.5-397b-fast	Qwen	262.13K	$0.6900/M	~1.17 Wh
kimi-k2.6	MoonshotAI	262.13K	$0.6900/M	~1.75 Wh
kimi-k2.6-fast	MoonshotAI	262.13K	$0.6900/M	~1.17 Wh
qwen3.6-35b	Qwen	131.06K	$0.2900/M	~1.56 Wh
qwen3.6-35b-fast	Qwen	131.06K	$0.2900/M	~0.97 Wh
kimi-k2.7-code	MoonshotAI	262.13K	$0.9500/M	~1.75 Wh
glm-5.2	ZhipuAI	1.05M	$1.45/M	~1.75 Wh
glm-5.2-fast	ZhipuAI	1.05M	$1.45/M	~1.17 Wh
glm-5.2-short	ZhipuAI	199.98K	$1.45/M	~1.36 Wh
glm-5.2-short-fast	ZhipuAI	199.98K	$1.45/M	~0.97 Wh