GLM 4.7 Flash Heretic
z-ai/glm-4.7-flash-heretic
Abliterated (Heretic) build of Zhipu GLM 4.7 Flash, fp8: the refusal direction is removed from the weights, reasoning and tool calling are kept. Thinking is on at low effort unless you set reasoning_effort (none/low/medium/high). Olafangensan's card identifies this as a modified GLM-4.7-Flash checkpoint built using Heretic 1.1.0. The underlying model is a 30B-total, 3B-active mixture of experts. Its published refusal measurements are checkpoint-author evaluations, not a guarantee for individual requests. Unlid declares a 131,072-token context and up to 24,000 output tokens. Input: text; output: text. Tool calling is supported. Reasoning is supported; it can be disabled. Reasoning tokens, when generated, count toward the output budget and are charged as output. Prices are USD per million tokens from the live catalog; missing cache discounts use the normal input price under Unlid billing rules. Limits and capabilities are provider declarations, not results of an inference test.
Model specifications
- Input
- text
- Output
- text
- Context
- 131,072 tokens
- Max output
- 24,000 tokens
- Input price
- $0.084 / 1M tokens
- Output price
- $0.48 / 1M tokens
- Released
- 2026-10-07
Capabilities
- Streaming
- Function calling
- Vision
- JSON mode
Provider pricing, discounts and data privacy
Compare effective provider prices, published discounts, regions, retention policies, training use, compliance, and privacy links by service tier.
Standard service tier
1 available provider · tier input average $0.084 / 1M tokens · tier output average $0.48 / 1M tokens
Unlid
Tier: Standard · Region: US · Quantization: fp8
Pricing
- Input
- $0.084 / 1M tokens
- Output
- $0.48 / 1M tokens
- Cache read
- $0.042 / 1M tokens
No provider discount is currently published.
Data privacy and compliance
- Region
- US
- Zero data retention
- No
- Used for training
- No
- Data collection
- Moderated
- No
- GDPR compliant
- No
- HIPAA compliant
- No
- SOC 2 certified
- No
- BYOK supported
- No
Frequently asked questions
- What is GLM 4.7 Flash Heretic?
- Abliterated (Heretic) build of Zhipu GLM 4.7 Flash, fp8: the refusal direction is removed from the weights, reasoning and tool calling are kept. Thinking is on at low effort unless you set reasoning_effort (none/low/medium/high). Olafangensan's card identifies this as a modified GLM-4.7-Flash checkpoint built using Heretic 1.1.0. The underlying model is a 30B-total, 3B-active mixture of experts. Its published refusal measurements are checkpoint-author evaluations, not a guarantee for individual requests. Unlid declares a 131,072-token context and up to 24,000 output tokens. Input: text; output: text. Tool calling is supported. Reasoning is supported; it can be disabled. Reasoning tokens, when generated, count toward the output budget and are charged as output. Prices are USD per million tokens from the live catalog; missing cache discounts use the normal input price under Unlid billing rules. Limits and capabilities are provider declarations, not results of an inference test.
- How much does GLM 4.7 Flash Heretic cost?
- Input costs start at $0.084 / 1M tokens and output costs start at $0.48 / 1M tokens. Provider-level prices vary by service tier.
- What is the context length of GLM 4.7 Flash Heretic?
- GLM 4.7 Flash Heretic supports a 131,072 token context window and up to 24,000 output tokens.
- What capabilities does GLM 4.7 Flash Heretic support?
- GLM 4.7 Flash Heretic supports Streaming, Function calling, Vision, JSON mode.
- Which providers offer GLM 4.7 Flash Heretic?
- GLM 4.7 Flash Heretic is available from Unlid.
- How do providers handle data privacy for GLM 4.7 Flash Heretic?
- 1 of 1 providers report that customer data is not used for training, and 0 offer zero-data-retention routing. Retention, compliance, and privacy-policy links are listed per provider.