Ask AI

New Models Added: Grok 4.20 Beta, GLM-5-Turbo, GPT-5.4 Mini/Nano, and MiniMax M2.7

Date: 2026-03-18 / (1404-12-27)

Summary

We announce the addition of seven new models: X.AI's Grok 4.20 Beta (reasoning and non-reasoning variants) with industry-leading speed and 2M context window, Z.AI's GLM-5-Turbo optimized for OpenClaw agentic scenarios, OpenAI's GPT-5.4 Mini and Nano for cost-efficient high-volume tasks, and MiniMax's M2.7 with self-evolution capabilities and SOTA coding performance.

Details

X.AI (Grok)

Grok 4.20 Beta Reasoning

grok-4.20-beta-0309-reasoning is X.AI's newest flagship model with industry-leading speed and agentic tool calling capabilities. It combines the lowest hallucination rate on the market with strict prompt adherence, delivering consistently precise and truthful responses.

Feature	Details
Context window	2,000,000 tokens
Input pricing	$2.00 / 1M tokens
Cached input pricing	$0.20 / 1M tokens (90% cost reduction)
Output pricing	$6.00 / 1M tokens
Input pricing (above 200K)	$4.00 / 1M tokens
Cached input (above 200K)	$0.40 / 1M tokens
Output pricing (above 200K)	$12.00 / 1M tokens
Input modalities	Text, Image
Output modalities	Text
Supported endpoints	`v1/chat/completions`

Key Features:

Reasoning Mode: The model thinks before responding for complex problem-solving
Massive Context: 2M token context window for extensive documents and conversations
Lowest Hallucination Rate: Industry-leading accuracy and truthfulness
Strict Prompt Adherence: Consistently precise responses
Function Calling: Connect the model to external tools and systems
Structured Outputs: Return responses in specific, organized formats

Aliases: grok-4.20-beta, grok-4.20-beta-0309, grok-4.20-beta-latest, grok-4.20-beta-latest-reasoning, grok-4.20-beta-reasoning

Grok 4.20 Beta Non-Reasoning

grok-4.20-beta-0309-non-reasoning is the non-reasoning variant optimized for fast responses without extended thinking, perfect for high-throughput applications.

Feature	Details
Context window	2,000,000 tokens
Input pricing	$2.00 / 1M tokens
Cached input pricing	$0.20 / 1M tokens (90% cost reduction)
Output pricing	$6.00 / 1M tokens
Input pricing (above 200K)	$4.00 / 1M tokens
Output pricing (above 200K)	$12.00 / 1M tokens
Input modalities	Text, Image
Output modalities	Text
Supported endpoints	`v1/chat/completions`

Key Features:

Fast Inference: Optimized for quick responses without reasoning overhead
Massive Context: 2M token context window
High-Throughput: Perfect for latency-sensitive applications
Function Calling & Structured Outputs: Full tool support

Z.AI (GLM)

GLM-5-Turbo

glm-5-turbo is Z.AI's foundation model deeply optimized for the OpenClaw agentic scenario. It has been specifically optimized for core requirements of OpenClaw tasks since the training phase, enhancing key capabilities such as tool invocation, command following, timed and persistent tasks, and long-chain execution.

Feature	Details
Context window	200,000 tokens
Maximum output	128,000 tokens
Input pricing	$1.32 / 1M tokens
Cached input pricing	$0.264 / 1M tokens (80% cost reduction)
Output pricing	$4.40 / 1M tokens
Input modalities	Text
Output modalities	Text
Supported endpoints	`v1/chat/completions`

Key Features:

OpenClaw Native Model: Systematically constructed for real-world agent workflows
Tool Calling: Precise invocation with greater stability and reliability in multi-step tasks
Instruction Following: Enhanced decomposition of complex, multi-layered instructions
Scheduled Tasks: Better understanding of time dimensions for scheduled triggers and long-running tasks
High-Throughput Long Chains: Faster and more stable execution for tasks with high data throughput
Thinking Mode: Multiple thinking modes for different scenarios
Streaming Output: Real-time streaming responses
Context Caching: Intelligent caching mechanism to optimize performance
Structured Output: Support for JSON and other structured formats
MCP Support: Flexibly integrate external MCP tools and data sources

OpenAI

GPT-5.4 Mini

gpt-5.4-mini is OpenAI's strongest mini model yet for coding, computer use, and subagents. It brings the strengths of GPT-5.4 to a faster, more efficient model designed for high-volume workloads.

Feature	Details
Context window	400,000 input tokens, 128,000 output tokens
Training data	Up to August 31, 2025
Input pricing	$0.75 / 1M tokens
Cached input pricing	$0.075 / 1M tokens (90% cost reduction)
Output pricing	$4.50 / 1M tokens
Input modalities	Text, Image
Output modalities	Text
Supported endpoints	`v1/chat/completions`, `v1/responses`

Key Features:

Reasoning Token Support: Higher reasoning capabilities with fast speed
400K Context Window: Extended conversations and document processing
128K Max Output Tokens: Extended generation capability
Function Calling & Structured Outputs: Full support for tools
Streaming: Full streaming support
Distillation Support: Can be used for model distillation
Computer Use: Supported via Responses API
MCP Support: Model Context Protocol support

Aliases: gpt-5.4-mini-2026-03-17

GPT-5.4 Nano

gpt-5.4-nano is OpenAI's cheapest GPT-5.4-class model for simple high-volume tasks. Designed for tasks where speed and cost matter most like classification, data extraction, ranking, and sub-agents.

Feature	Details
Context window	400,000 input tokens, 128,000 output tokens
Training data	Up to August 31, 2025
Input pricing	$0.20 / 1M tokens
Cached input pricing	$0.02 / 1M tokens (90% cost reduction)
Output pricing	$1.25 / 1M tokens
Input modalities	Text, Image
Output modalities	Text
Supported endpoints	`v1/chat/completions`, `v1/responses`

Key Features:

High Reasoning: Strong reasoning at the lowest cost
Fast Speed: Optimized for high-volume tasks
400K Context Window: Extended context handling
128K Max Output Tokens: Extended generation
Function Calling & Structured Outputs: Full support
Streaming: Full streaming support
MCP Support: Model Context Protocol support

Aliases: gpt-5.4-nano-2026-03-17

MiniMax

MiniMax M2.7

minimax-m2.7 is MiniMax's first model deeply participating in its own evolution. M2.7 is capable of building complex agent harnesses and completing highly elaborate productivity tasks, leveraging capabilities such as Agent Teams, complex Skills, and dynamic tool search.

Feature	Details
Context window	204,000 tokens
Maximum output	128,000 tokens
Input pricing	$0.30 / 1M tokens
Cached input pricing	$0.06 / 1M tokens (80% cost reduction)
Cache creation pricing	$0.375 / 1M tokens
Output pricing	$1.20 / 1M tokens
Input modalities	Text
Output modalities	Text
Supported endpoints	`v1/chat/completions`

Key Features:

Self-Evolution: First model that deeply participates in its own evolution
SOTA Software Engineering: 56.22% on SWE-Pro, nearly matching Opus's best level
Professional Productivity: ELO score of 1495 on GDPval-AA (highest among open-source models)
Complex Environment Interaction: 97% skill adherence rate with 40+ complex skills
Agent Teams: Native support for multi-agent collaboration
Office Suite Integration: Enhanced Word, Excel, PPT handling with multi-round high-fidelity editing
Character Consistency: Excellent emotional intelligence for interactive entertainment
Output Speed: Approximately 60 tokens per second

Performance Highlights:

End-to-end full project delivery (VIBE-Pro 55.6%)
Terminal Bench 2 (57.0%) for deep understanding of complex engineering systems
Multi-SWE-Bench (52.7%)
SWE Multilingual (76.5%)

MiniMax M2.7 Highspeed

minimax-m2.7-highspeed is the ultra-fast variant of M2.7, optimized for speed while maintaining identical capabilities.

Feature	Details
Context window	204,000 tokens
Maximum output	128,000 tokens
Input pricing	$0.60 / 1M tokens
Cached input pricing	$0.06 / 1M tokens
Cache creation pricing	$0.375 / 1M tokens
Output pricing	$2.40 / 1M tokens
Input modalities	Text
Output modalities	Text
Supported endpoints	`v1/chat/completions`

Key Features:

Output Speed: Approximately 100 tokens per second (faster than other frontier models)
Same Capabilities as M2.7: Identical intelligence with higher throughput
Low-Cost Operation: Efficient for real-time applications
Full Tool Support: Same tool calling and skill support as M2.7

Pricing Summary

Model	Input ($/1M tokens)	Cached Input ($/1M tokens)	Output ($/1M tokens)
`grok-4.20-beta-0309-reasoning`	$2.00	$0.20	$6.00
`grok-4.20-beta-0309-non-reasoning`	$2.00	$0.20	$6.00
`glm-5-turbo`	$1.32	$0.264	$4.40
`gpt-5.4-mini`	$0.75	$0.075	$4.50
`gpt-5.4-nano`	$0.20	$0.02	$1.25
`minimax-m2.7`	$0.30	$0.06	$1.20
`minimax-m2.7-highspeed`	$0.60	$0.06	$2.40

API Request/Response Examples

Grok 4.20 Beta Reasoning Example

bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "grok-4.20-beta-0309-reasoning",
    "messages": [
      {
        "role": "user",
        "content": "Design a comprehensive digital marketing strategy for a tech startup."
      }
    ],
    "max_tokens": 4096
  }'

GLM-5-Turbo with Thinking Mode

bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "glm-5-turbo",
    "messages": [
      {
        "role": "user",
        "content": "As a marketing expert, create an attractive slogan for Z.AI Open Platform."
      }
    ],
    "thinking": {
      "type": "enabled"
    },
    "max_tokens": 4096,
    "temperature": 1.0
  }'

GPT-5.4 Mini Example

bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "gpt-5.4-mini",
    "messages": [
      {
        "role": "user",
        "content": "Implement a rate-limited API client in Python with exponential backoff."
      }
    ],
    "max_tokens": 4096
  }'

GPT-5.4 Nano Example

bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "gpt-5.4-nano",
    "messages": [
      {
        "role": "user",
        "content": "Classify the following text as positive, negative, or neutral: 'The product works well but shipping was slow.'"
      }
    ]
  }'

MiniMax M2.7 Example

bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "minimax-m2.7",
    "messages": [
      {
        "role": "user",
        "content": "Build a research agent harness that can handle data pipelines, training environments, and cross-team collaboration."
      }
    ],
    "max_tokens": 8192
  }'

SDK Usage Examples

BashPythonJavaScript

bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "gpt-5.4-mini",
    "messages": [
      {
        "role": "user",
        "content": "Explain the key differences between microservices and monolithic architecture."
      }
    ]
  }'

python

from openai import OpenAI

client = OpenAI(api_key="your-avalai-api-key", base_url="https://api.avalai.ir/v1")

# Using GPT-5.4 Mini
response = client.chat.completions.create(
    model="gpt-5.4-mini",
    messages=[
        {
            "role": "user",
            "content": "Explain the key differences between microservices and monolithic architecture.",
        }
    ],
)

print(response.choices[0].message.content)

# Using Grok 4.20 Beta
response = client.chat.completions.create(
    model="grok-4.20-beta-0309-reasoning",
    messages=[
        {
            "role": "user",
            "content": "Design a fault-tolerant distributed system architecture.",
        }
    ],
    max_tokens=4096,
)

print(response.choices[0].message.content)

javascript

import { OpenAI } from "openai";

const client = new OpenAI({
  apiKey: process.env.AVALAI_API_KEY,
  baseURL: "https://api.avalai.ir/v1",
});

// Using GPT-5.4 Mini
const response = await client.chat.completions.create({
  model: "gpt-5.4-mini",
  messages: [
    {
      role: "user",
      content: "Explain the key differences between microservices and monolithic architecture.",
    },
  ],
});

console.log(response.choices[0].message.content);

// Using MiniMax M2.7
const m2Response = await client.chat.completions.create({
  model: "minimax-m2.7",
  messages: [
    {
      role: "user",
      content: "Build an agent that can autonomously manage ML experiments.",
    },
  ],
  max_tokens: 8192,
});

console.log(m2Response.choices[0].message.content);

New Models Added: Grok 4.20 Beta, GLM-5-Turbo, GPT-5.4 Mini/Nano, and MiniMax M2.7 ​

Summary ​

Details ​

X.AI (Grok) ​

Grok 4.20 Beta Reasoning ​

Grok 4.20 Beta Non-Reasoning ​

Z.AI (GLM) ​

GLM-5-Turbo ​

OpenAI ​

GPT-5.4 Mini ​

GPT-5.4 Nano ​

MiniMax ​

MiniMax M2.7 ​

MiniMax M2.7 Highspeed ​

Pricing Summary ​

API Request/Response Examples ​

Grok 4.20 Beta Reasoning Example ​

GLM-5-Turbo with Thinking Mode ​

GPT-5.4 Mini Example ​

GPT-5.4 Nano Example ​

MiniMax M2.7 Example ​

SDK Usage Examples ​

Documentation Links ​

New Models Added: Grok 4.20 Beta, GLM-5-Turbo, GPT-5.4 Mini/Nano, and MiniMax M2.7

Summary

Details

X.AI (Grok)

Grok 4.20 Beta Reasoning

Grok 4.20 Beta Non-Reasoning

Z.AI (GLM)

GLM-5-Turbo

OpenAI

GPT-5.4 Mini

GPT-5.4 Nano

MiniMax

MiniMax M2.7

MiniMax M2.7 Highspeed

Pricing Summary

API Request/Response Examples

Grok 4.20 Beta Reasoning Example

GLM-5-Turbo with Thinking Mode

GPT-5.4 Mini Example

GPT-5.4 Nano Example

MiniMax M2.7 Example

SDK Usage Examples

Documentation Links